Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Applicant’s arguments in view of the claim amendments filed on 12/30/2025, with respect to the 35 U.S.C. §112 rejections have been fully considered and are persuasive. The 35 U.S.C. §112 rejections have been withdrawn.
Applicant’s arguments in view of the claim amendments filed on 12/30/2025, with respect to the prior art rejections have been fully considered and are persuasive. Therefore, the rejections have been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Achin et al., Showalter, Lundberg et al. and Drinkwater.
Applicant's arguments in view of the claim amendments filed on 12/30/2025, with regard to the double patenting rejection, have been fully considered but they are not fully persuasive. Because the instant claims and the reference patent applications claims’ have been amended since the filing of the Non-Final Office Action, the obvious double patent rejection has been updated accordingly with corresponding obviousness analysis.
Applicant's arguments filed 12/30/2025 with respect to the rejection of claims 1-20 under 35 U.S.C. § 101 have been fully considered but they are not persuasive. The
rejection of claims 1-20 under 35 U.S.C. § 101 is maintained for the reasons of
record set forth at pages 31-58 of the Office action mailed 09/30/2025, as
supplemented by the rebuttal set forth below and by the analysis set forth in the 35 U.S.C. 101 rejection that follows for the amended claims.
Applicant states that the independent claims are eligible "even in unamended form" (Remarks, pp. 18, 20). The rebuttal below is accordingly directed to the grounds of rejection already of record and to the arguments Applicant has raised against them.
To the extent the analysis below addresses the limitation "the explainability data
characterizing a contribution of at least one feature value of the input dataset to
the portion of the output data," newly added to claims 1, 12, and 20 by the amendment
filed 12/30/2025, that portion of the rejection constitutes a new ground necessitated
by Applicant's amendment. See MPEP § 706.07(a).
I. Step 2A, Prong One
Applicant argues (Remarks, pp. 16-17) that the Office failed to identify the specific limitations believed to recite an abstract idea, as required by MPEP § 2106.04(a). This argument is not persuasive. The limitations of each claim believed to recite the judicial exception are identified verbatim, claim by claim, in the rejection set forth below, and each identified limitation is matched to the "mental processes" grouping of MPEP § 2106.04(a)(2)(III).
Applicant argues (Remarks, p. 17) that the Office failed to afford the claims their broadest reasonable interpretation consistent with the specification, as required by MPEP § 2111, and that no portion of the specification supports the conclusion that a person could perform the identified limitations mentally or with pen and paper. This argument is not persuasive. The Prong One characterization set forth below rests on what the claims themselves recite and does not import any limitation from the specification. The claims place no bound on the size of the input dataset, on the number of features it contains, or on the complexity of the predicted likelihood. Compiling a set of values from prior interaction records, forming a judgment as to how likely a future event is, and attributing that judgment to particular compiled factors are observations, evaluations, and judgments of the kind expressly enumerated in MPEP § 2106.04(a)(2)(III). See Electric Power Group v. Alstom, S.A., 830 F.3d 1350, 1353-54 (Fed. Cir. 2016) (claim to "collecting information, analyzing it, and displaying certain results of the collection and analysis," where the data analysis steps are recited at a high level of generality, recites a mental process), cited at MPEP § 2106.04(a)(2)(III).
Applicant argues (Remarks, p. 17) that the Office improperly "abstracts out" the
claimed "trained artificial intelligence process" from the recited elements, and (Remarks, p. 18) that the claimed combination encompasses artificial intelligence in a manner that cannot practically be performed in the human mind, citing MPEP § 2106.04(a)(2)(III)(A) and page 2 of the memorandum of August 4, 2025. These arguments are not persuasive, because the rejection does not do what Applicant describes. Consistent with the August 4, 2025 memorandum, the recitation of "a trained artificial intelligence process" is NOT treated as part of the judicial exception and is NOT characterized as a mental process. It is treated as an additional element and is analyzed as such at Prong Two and at Step 2B below. The Office agrees with Applicant that "[c]laims do not recite a mental process when they do not contain limitations that can practically be performed in the human mind, for instance when the human mind is not equipped to perform the claim limitations" (MPEP § 2106.04(a)(2)(III)(A)), and the mental-process grouping has not been extended to any such limitation here. The grouping reaches only the underlying acts of compiling information, forming a judgment as to likelihood, and attributing that judgment to particular factors — acts that are recited without any bound that would place them beyond the human mind.
II. Step 2A, Prong Two
Applicant argues (Remarks, pp. 19-20) that the Office evaluated the additional elements in a vacuum and failed to consider the claim as a whole, citing MPEP § 2106.04(d)(II) and pages 3-4 of the August 4, 2025 memorandum. This argument is not persuasive. The Prong Two analysis set forth below evaluates the additional elements individually and as an ordered combination, and evaluates them together with the limitations reciting the judicial exception, in accordance with MPEP § 2106.04(d)(II). The conclusion does not rest on any element considered in isolation, and no additional element has been disregarded on the ground that it is conventional. See MPEP § 2106.04(d).
Applicant argues (Remarks, p. 20) that Ex Parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025) (precedential), cautions that examiners "should not evaluate claims at such a high level of generality" at Prong Two. Applicant is correct that this is the governing standard, and it is followed here. That language now appears in MPEP § 2106.05(a) as revised by the memorandum of December 5, 2025, "Advance notice of change to the MPEP in light of Ex Parte Desjardins," which is effective upon issuance and supersedes the corresponding text of the Ninth Edition, Revision 01.2024. The revised section directs that the claim be evaluated "as an ordered combination, without ignoring the requirements of the individual steps," that examiners be "careful to avoid oversimplifying the claims," and that examiners "should not evaluate claims at such a high level of generality" that potentially meaningful technical limitations are dismissed without adequate explanation.
The rejection conforms to that standard. Consistent with MPEP § 2106.04(d)(1) as revised by the same memorandum, the specification was consulted first to determine whether the disclosure provides sufficient detail that one of ordinary skill in the art would recognize the claimed invention as providing an improvement in the functioning of a computer or in another technology or technical field; the specific improvement the specification describes was identified; and each independent claim was then evaluated to determine whether the claim itself reflects that disclosed improvement — that is, whether the claim includes the components or steps of the invention that provide it. No potentially meaningful technical limitation has been dismissed without explanation, and each limitation Applicant identifies is addressed on its own terms below. The rejection does not rest on a characterization of the claims at a high level of generality. It rests on the specific finding, set forth below, that the parallelized, distributed, GPU/TPU-based implementation to which the specification attributes the asserted improvement appears nowhere in the claims.
Applicant argues (Remarks, pp. 20-22) that the claims recite a "specific, technological improvement" to existing computer-implemented predictive processes that "reduces a number of discrete computational operations, and an amount of computational resources," citing Specification ¶¶ [0015]-[0022], Koninklijke KPN N.V. v. Gemalto M2M GmbH, 942 F.3d 1143 (Fed. Cir. 2019), and Enfish, LLC v. Microsoft Corp., 822 F.3d 1327 (Fed. Cir. 2016). Because Applicant has placed the cited specification passages directly at issue, those passages have been examined. They do not support the argument. The specification attributes the asserted real-time capability to "an implementation of one or more parallelized, fault-tolerant distributed computing and analytical protocols across clusters of graphical processing units (GPUs) and/or
tensor processing units (TPUs)" (¶ [022]), to "an Apache Spark™ distributed, cluster-computing framework, a Databricks™ analytical platform," and to storage
"within a portion of a distributed file system, such as a Hadoop distributed file
system (HDFS)" (¶ [026]).
None of that appears in the claims. Claim 1 recites no parallelized processing, no distributed computing cluster or component, no graphics processing unit, no tensor processing unit, no cluster-computing framework, and no distributed file system. Claim 1 requires only "a memory," "a communications interface," and "at least one
processor" — and the specification itself describes such a processor as "a central processing unit (CPU) capable of processing a single operation (e.g., a scalar operation) in a single clock cycle" (¶ [024]) and as one of the "general or special purpose microprocessors" suitable for executing any computer program (
¶ [0158]). Under the broadest reasonable interpretation, claim 1 is fully satisfied by a single conventional processor performing the recited operations serially — precisely the configuration the specification identifies as inadequate. The Office does not rest this finding on the absence of any express recitation of the asserted benefit. As MPEP § 2106.04(d)(1) provides, "[t]he claim itself does not need to explicitly recite the improvement described in the specification (e.g., 'thereby increasing the bandwidth of the channel')." The deficiency is not that claim 1 omits the words "reduces computational operations"; it is that claim 1 omits the components and steps that produce that reduction. The revised section requires that "the claim must include the components or steps of the invention that provide the improvement described in the specification," and MPEP § 2106.05(a) as revised makes the controlling comparison explicit by citation to Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1316 (Fed. Cir. 2016), in which "patent owner argued that the claimed email filtering system improved technology by shrinking the protection gap and mooting the volume problem, but the court disagreed because the claims themselves did not have any limitations that addressed these issues." That is this case. Claims 12 and 20 recite no such components or steps either, and the same analysis applies to each.
The distinction drawn in Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025), cited by Applicant at Remarks p. 22, is to the same effect: steps incidental to automating an abstract idea are not sufficient to confer eligibility. Applicant's claims apply an already-trained artificial intelligence process to a new body of interaction data in order to forecast events; they do not improve the machine learning
technology itself.
The claims are further distinguishable from those held eligible in Desjardins. There, the specification identified an improvement to how the machine learning model itself operates — training a model to learn new tasks while protecting knowledge about previously learned tasks, so as to overcome "catastrophic forgetting" — and the claim recited the very limitation that delivered that improvement. The instant claims recite nothing comparable. The trained artificial intelligence process appears in claim 1 only as a black box: it is already trained when the claim begins, it is applied to an input dataset, and it emits output data. No limitation of claim 1 alters the structure of the model, the manner in which it is trained, the manner in which it is executed, its storage footprint, or its computational complexity. The model is invoked as a tool to carry out the recited prediction; it is not itself improved. See MPEP § 2106.05(a), (f). Koninklijke KPN and Enfish are distinguishable for the same reason: in each, the claim itself recited the specific mechanism that produced the asserted technical benefit.
III. Step 2B
Applicant argues (Remarks, p. 23) that the Office applied the same reasoning at Prong Two and at Step 2B and therefore failed to analyze Step 2B properly. This argument is not persuasive as applied to the analysis set forth below. Substantial overlap between the two inquiries is expected and proper: "Although most of these considerations overlap (i.e., they are evaluated in both Step 2A Prong Two and Step 2B), Step 2A specifically excludes consideration of whether the additional elements represent well-understood, routine, conventional activity." MPEP § 2106.04(d). The Step 2B analysis below performs precisely the inquiry that Prong Two excludes: each additional element is separately evaluated for whether it is well-understood, routine, and
conventional, and each such finding is supported by a citation to a court decision recognizing the element as well-understood, routine, and conventional, in accordance with MPEP § 2106.05(d)(II). Consistent with MPEP § 2106.05(a) as revised by the memorandum of December 5, 2025, no additional element has been dismissed as a mere "generic computer component" without considering whether it confers a technological improvement on a technical problem; each additional element has been evaluated for that possibility, and the reasons it does not are stated.
Applicant argues (Remarks, pp. 22-23) that the claimed elements, taken collectively, extend beyond well-understood, routine, or conventional activity. This argument is not persuasive for the reasons set forth in the Step 2B analysis below, which addresses the additional elements both individually and as an ordered combination and finds that they are arranged in the conventional order in which any computer performs any data analysis: input data is assembled, a computation is performed upon it, and the result is transmitted elsewhere. The claims recite no non-conventional or non-generic arrangement of the kind found in BASCOM Global Internet Services, Inc. v. AT&T Mobility LLC, 827 F.3d 1341, 1350 (Fed. Cir. 2016), or DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1258-59 (Fed. Cir. 2014).
Applicant argues (Remarks, pp. 23-24) that the claims recite an inventive concept under Step 2B because they are neither anticipated by nor rendered obvious over Dorai, Hollins, and the other references of record, citing Alice Corp. v. CLS Bank International, 573 U.S. 208, 217-18, 224-25 (2014). This argument is not persuasive
for two independent reasons. First, novelty is not the test for eligibility. "[T]he search for an inventive concept should not be confused with a novelty or non-obviousness determination," and the "'novelty' of any element or steps in a process, or even of the process itself, is of no relevance in determining whether the subject matter of a claim falls within the § 101 categories of possibly patentable subject matter." MPEP § 2106.05, quoting Intellectual Ventures I v. Symantec Corp., 838 F.3d 1307, 1315 (Fed. Cir. 2016), and Diamond v. Diehr, 450 U.S. 175, 188-89 (1981); see also Synopsys, Inc. v. Mentor Graphics Corp., 839 F.3d 1138, 1151 (Fed. Cir. 2016) ("a claim for a new abstract idea is still an abstract idea. The search for a § 101 inventive concept is thus distinct from demonstrating § 102 novelty."). An inventive concept "cannot be furnished by the unpatentable law of nature (or natural phenomenon or abstract idea) itself." MPEP § 2106.05, quoting Genetic Techs. Ltd. v. Merial LLC, 818 F.3d 1369, 1376 (Fed. Cir. 2016). Second, the factual premise of the argument no longer obtains. The rejections over Dorai and Hollins have been withdrawn, and claims 1-20 stand rejected under 35 U.S.C. § 103 over Achin, Showalter, Lundberg, and Drinkwater as set forth below.
For the reasons set forth above and in the rejection that follows, the rejection of
claims 1-20 under 35 U.S.C. § 101 is maintained.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis below follows the 2019 Revised Patent Subject Matter Eligibility Guidance as codified in MPEP 2106, the precedential Appeals Review Panel decision in Ex Parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025), and the August 4, 2025 memorandum to Technology Centers 2100, 2600, and 3600 on evaluating subject matter eligibility under 35 U.S.C. 101.
CLAIM 1
Step 1: Claim 1 recites “An apparatus, comprising: a memory storing instructions; a communications interface; and at least one processor coupled to the memory and the communications interface …”, that is, a combination of physical components. Claim 1 therefore falls within the statutory category of a machine. See MPEP 2106.03. Accordingly, claim 1 satisfies Step 1 of the eligibility analysis.
Step 2A, Prong 1: Claim 1 recites a judicial exception, namely an abstract idea falling within the “certain methods of organizing human activity” and “mental processes” groupings. See MPEP 2106.04(a)(2)(II) and (III). The limitations reciting the judicial exception are: “generate an input dataset … based on elements of first interaction data associated with a first temporal interval”; “generate output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval”; and “the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data”.
Set apart from the recited computer components and the generically recited artificial intelligence process, these limitations describe the practice of compiling a record of a party’s prior interactions over a past window, forecasting how likely it is that specified events involving that party will occur during a later window offset from the past window, and identifying which of the compiled factors drove that forecast. As the specification confirms, the events forecast are redemptions of a financial product held by a customer of a financial institution, and the object of the forecast is to allow the institution to “engage, proactively, one or more of the customers … in an attempt to prevent any future redemption” and to “maintain the customers’ positions in the mutual fund products” (specification ¶[017], ¶[021]). Forecasting a customer’s future disposition of a financial product in order to intervene commercially and retain that customer’s business is a fundamental economic practice and a commercial interaction, and thus falls within the “certain methods of organizing human activity” grouping. See MPEP 2106.04(a)(2)(II); Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208 (2014); Bilski v. Kappos, 561 U.S. 593 (2010). This characterization rests on what claim 1 itself recites, and not on any limitation imported from the specification. See MPEP 2111. Claim 1 recites “interaction data” associated with a party, the forecasting of “targeted events” involving that party during a later interval, and a “computing system … configured to perform operations based on” that forecast. Under the broadest reasonable interpretation, that language covers the management of a commercial relationship between an institution and its counterparties — forecasting a counterparty's future conduct so that the institution may act upon it. The specification's disclosure that the targeted events are mutual fund redemptions and that the operations performed are proactive customer engagement confirms this reading but is not relied upon to supply it; claims 8-11 and 18-19 expressly recite that redemption embodiment.
These same limitations independently recite a mental process, on a basis that does not depend on the recited artificial intelligence process. See MPEP 2106.04(a)(2)(III). Compiling a set of values from prior interaction records, forming a judgment as to how likely a future event is, and assessing which of the compiled values most influenced that judgment are observations, evaluations, and judgments that a person can practically perform in the mind or with pen and paper. The claim places no bound on the size of the input dataset, the number of features it contains, or the complexity of the forecast. Consistent with the August 4, 2025 memorandum, this grouping is expressly NOT extended to the act of generating output data by applying the trained artificial intelligence process, which is analyzed below as an additional element; the mental-process basis reaches only the underlying acts of compiling information, forming a judgment, and attributing that judgment to particular factors. The two groupings identified above are independent bases, and either alone establishes that claim 1 recites a judicial exception.
Consistent with the August 4, 2025 memorandum on evaluating subject matter eligibility, the recitation of “a trained artificial intelligence process” is not treated as part of the judicial exception. That limitation is recited generically and does not set forth or describe any mathematical relationship, calculation, formula, or equation using words or mathematical symbols; it therefore merely involves, rather than recites, a mathematical concept. Compare USPTO Subject Matter Eligibility Example 39 (the generic limitation “training the neural network in a first stage using the first training set” does not recite a judicial exception) with Example 47, claim 2 (training “wherein the selected training algorithm includes a backpropagation algorithm and a gradient descent algorithm” does recite one, because specific calculations are named). See MPEP 2106.04, subsection II(A)(1); Ex Parte Linden, No. 2018-003323 (PTAB Apr. 1, 2019).
Nor is the application of the artificial intelligence process characterized as a mental process; the mental-process grouping is not expanded to reach limitations that cannot practically be performed in the human mind. The artificial intelligence process is instead treated as an additional element and analyzed below.
Step 2A, Prong 2: The additional elements of claim 1 are: “a memory storing instructions”; “a communications interface”; “at least one processor coupled to the memory and the communications interface, the at least one processor being configured to execute the instructions to”; “a trained artificial intelligence process” and the “application of the trained artificial intelligence process to the input dataset”; “transmit at least a portion of the output data and explainability data associated with the trained artificial intelligence process to a computing system via the communications interface”; and “the computing system being configured to perform operations based on the portion of the output data and the explainability data”.
The judicial exception is not integrated into a practical application. The additional elements, considered individually and as an ordered combination, do not integrate the judicial exception into a practical application. Per Ex Parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025), and MPEP 2106.04(d)(1), the specification has been consulted first to determine whether the disclosed invention improves the functioning of a computer or another technology or technical field. The specification does describe subject matter that a person of ordinary skill in the art could recognize as a technological improvement, but the improvement described is one of data-processing throughput achieved through a parallelized, distributed hardware implementation. The specification states that “given the increasing volume of the data … some existing analytical processes may be incapable of analyzing the maintained elements of customer- or interaction-specific data in time frames sufficient to support a determination that a customer represents a likely candidate for involvement in a redemption event in real-time” (specification ¶[018]), and attributes the resulting real-time capability to “an implementation of one or more parallelized, fault-tolerant distributed computing and analytical protocols across clusters of graphical processing units (GPUs) and/or tensor processing units (TPUs)” (¶[022]), to “an Apache Spark™ distributed, cluster-computing framework, a Databricks™ analytical platform,” and to storage “within a portion of a distributed file system, such as a Hadoop distributed file system (HDFS)” (¶[026]). The remaining benefits the specification asserts — identifying customers likely to redeem, engaging them proactively, and preventing future redemptions so as to “maintain the customers’ positions in the mutual fund products” (¶[017]; ¶[021]) — are commercial benefits to the financial institution, not improvements to computer functionality or to any other technology or technical field.
The claim, however, does not reflect the improvement disclosed in the specification. Claim 1 recites no parallelized processing, no distributed computing cluster or distributed computing component, no graphics processing unit, no tensor processing unit, no cluster-computing framework, and no distributed file system. Claim 1 requires only “a memory,” “a communications interface,” and “at least one processor” — and the specification itself describes such a processor as “a central processing unit (CPU) capable of processing a single operation (e.g., a scalar operation) in a single clock cycle” (¶[024]) and as one of the “general or special purpose microprocessors” suitable for executing any computer program (¶[0158]). Under the broadest reasonable interpretation, claim 1 is fully satisfied by a single conventional processor performing the recited operations serially — precisely the configuration the specification identifies as inadequate. Because the claim does not include the components that provide the improvement described in the specification, the improvement consideration does not apply. See MPEP 2106.05(a); Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1316 (Fed. Cir. 2016) (claims did not reflect the alleged improvement because they contained no limitations addressing it).
Claim 1 is further distinguishable from the claims held eligible in Ex Parte Desjardins. There, the specification identified an improvement to how the machine learning model itself operates — training a model to learn new tasks while protecting knowledge about previously learned tasks so as to overcome the problem of “catastrophic forgetting” in continual learning systems — and the claim recited the very limitation that delivered that improvement, namely adjusting parameter values “to optimize performance of the machine learning model on the second machine learning task while protecting performance of the machine learning model on the first machine learning task.” Claim 1 recites nothing comparable. The trained artificial intelligence process appears in claim 1 only as a black box: it is already trained when the claim begins, it is applied to an input dataset, and it emits output data. No limitation of claim 1 alters the structure of the model, the manner in which it is trained, the manner in which it is executed, its storage footprint, or its computational complexity. The model is invoked as a tool to carry out the recited prediction; it is not itself improved. See MPEP 2106.05(a), (f).
The buffer-interval limitation does not supply a technological improvement. “the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval” specifies only which data is drawn upon and when the forecast applies. The specification describes the buffer interval as having “a predetermined duration, such as, but not limited to, one month,” established “to separate temporally the customers’ prior interactions with the financial institution … from the future target temporal interval” (¶[056]). No mechanism is recited, and no technical effect on the operation of any computer or model is described. Selecting a particular data source, or a particular type of data to be manipulated, is insignificant extra-solution activity. See MPEP 2106.05(g). Moreover, the temporal relationship between the two intervals is part of the recited abstract idea itself — it defines the forecast — and a limitation that is part of the judicial exception cannot integrate that exception into a practical application. See MPEP 2106.04(d)(2).
The explainability limitation likewise does not supply a technological improvement. “the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data” recites the result to be obtained — data that characterizes a feature’s contribution — without reciting any particular way of obtaining it. The specification discloses specific mechanisms for computing such a contribution, including “a determined number of branching points that utilize the corresponding feature” and “a computed Shapley feature value for the corresponding feature” (¶[083]), but claim 1 recites neither, and instead encompasses any means of characterizing any feature’s contribution to the output. A claim that recites only the idea of a solution or outcome, rather than a particular way to achieve that outcome, does not improve technology or a technical field. See MPEP 2106.05(a); McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1314-15 (Fed. Cir. 2016). In any event, the assessment of which feature contributed to a conclusion is itself part of the recited abstract idea, and is therefore not an additional element available to integrate the exception. See MPEP 2106.04(d)(2).
Turning to the additional elements individually: “a memory storing instructions,” “a communications interface,” and “at least one processor coupled to the memory and the communications interface, the at least one processor being configured to execute the instructions to” are recited at a high level of generality and amount to no more than generic computer components invoked in their ordinary capacities as a tool to perform the abstract idea. Mere recitation that a judicial exception is to be performed using generic computer equipment in its ordinary capacity cannot meaningfully integrate the exception into a practical application. See MPEP 2106.05(f). The specification confirms the genericity of these components, describing the “apparatus” as encompassing “all kinds of apparatus, devices, and machines for processing data” (¶[0155]) and stating that “[t]he essential elements of a computer are a central processing unit for performing or executing instructions and one or more memory devices for storing instructions and data” (¶[0158]).
The “trained artificial intelligence process” and the “application of the trained artificial intelligence process to the input dataset” are likewise recited at a high level of generality. The claim identifies no architecture, no training regime, no parameters, and no computation; it recites only that some trained artificial intelligence process is applied to a dataset and that output data results. This is the use of a generic class of computer algorithm in its ordinary capacity to perform the abstract idea, and amounts to the equivalent of an instruction to “apply it.” See MPEP 2106.05(f).
The limitation “transmit at least a portion of the output data and explainability data associated with the trained artificial intelligence process to a computing system via the communications interface” is post-solution output activity: it conveys the result of the abstract idea to another location after the idea has been performed. Transmitting the results of an analysis is insignificant extra-solution activity that does not integrate the exception into a practical application. See MPEP 2106.05(g).
The limitation “the computing system being configured to perform operations based on the portion of the output data and the explainability data” recites no particular operation whatever. It is a statement of the intended use of a generic recipient computer and, at most, generally links the abstract idea to a computer environment. See MPEP 2106.05(f), (h).
Considered as a whole and as an ordered combination, the additional elements of claim 1 do no more than gather data, apply an unspecified artificial intelligence process to it on a generic processor, and transmit the result over a generic interface. The claim as a whole does not improve the functioning of a computer or any other technology or technical field. Accordingly, claim 1 does not satisfy Step 2A, Prong 2.
The remaining integration considerations do not apply. Claim 1 does not tie the judicial exception to a particular machine: the recited memory, communications interface, and processor are claimed without any particular structure, arrangement, or configuration, and a generic computer is not a particular machine that imposes meaningful limits. See MPEP 2106.05(b). Nor does claim 1 effect a transformation or reduction of a particular article to a different state or thing — the claim manipulates data only, and generating output data from input data is not a transformation of matter. See MPEP 2106.05(c). No other meaningful limitation is present. See MPEP 2106.05(e).
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. Each additional element is well-understood, routine, and conventional. The “memory storing instructions” and the storage of data therein amount to storing and retrieving information in memory and to electronic recordkeeping, which the courts have recognized as well-understood, routine, and conventional. See Versata Development Group, Inc. v. SAP America, Inc., 793 F.3d 1306, 1334 (Fed. Cir. 2015); OIP Technologies, Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363 (Fed. Cir. 2015); Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 225-26 (2014); MPEP 2106.05(d)(II). The “at least one processor … configured to execute the instructions” performs the generic computer functions of executing stored instructions and carrying out the recited operations. This element has been considered for whether it confers a technological improvement on a technical problem, and it does not: claim 1 specifies no architecture, configuration, or manner of operation for the processor beyond its coupling to the memory and the communications interface, and the specification describes such a processor as “a central processing unit (CPU) capable of processing a single operation (e.g., a scalar operation) in a single clock cycle” (¶[024]) and as one of the “general or special purpose microprocessors” suitable for executing any computer program (¶[0158]). The processor is therefore invoked as a tool to carry out the recited prediction rather than as a component that is itself improved. See Alice Corp., 573 U.S. at 225-26; MPEP 2106.05(a) and (f); MPEP 2106.05(d)(II). The “communications interface” and the transmission of output data and explainability data to a computing system amount to receiving or transmitting data over a network, which the courts have likewise recognized as well-understood, routine, and conventional. See Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1321 (Fed. Cir. 2016); TLI Communications LLC v. AV Automotive LLC, 823 F.3d 607, 610 (Fed. Cir. 2016); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355 (Fed. Cir. 2014); MPEP 2106.05(d)(II). The “trained artificial intelligence process” and its “application … to the input dataset” are, as recited, the performance of repetitive calculations by a computer, which the courts have found to be well-understood, routine, and conventional. See Parker v. Flook, 437 U.S. 584, 594 (1978); Bancorp Servs., L.L.C. v. Sun Life Assurance Co. of Canada (U.S.), 687 F.3d 1266, 1278 (Fed. Cir. 2012); MPEP 2106.05(d)(II). The specification confirms this by identifying the process as an off-the-shelf XGBoost gradient-boosted decision-tree process executed on commercially available distributed-computing platforms (¶[019]; ¶[026]).
The specification independently establishes the well-understood, routine, and conventional nature of these elements, which is Berkheimer-compliant evidence of the first type. See Berkheimer v. HP Inc., 881 F.3d 1360 (Fed. Cir. 2018); MPEP 2106.05(d)(I). The specification states that the disclosed functional operations “can be implemented in digital electronic circuitry, in tangibly-embodied computer software or firmware, in computer hardware, including the structures disclosed in this specification and their structural equivalents” (¶[0153]); that “[c]omputers suitable for the execution of a computer program include, by way of example, general or special purpose microprocessors or both, or any other kind of central processing unit,” and that “[t]he essential elements of a computer are a central processing unit for performing or executing instructions and one or more memory devices for storing instructions and data” (¶[0158]); and that the components “can be interconnected by any form or medium of digital data communication, such as a communication network,” examples of which “include a local area network (LAN) and a wide area network (WAN), such as the Internet” (¶[0161]). The specification further identifies the artificial intelligence process as a known, commercially available technique — “a gradient-boosted decision-tree process (e.g., XGBoost process)” (¶[019]) — implemented on commercially available platforms, namely “an Apache Spark™ distributed, cluster-computing framework, a Databricks™ analytical platform” and “a Hadoop distributed file system (HDFS)” (¶[026]).
Considered as an ordered combination, the additional elements add nothing that is not already present when they are considered individually. They are arranged in the conventional order in which any computer performs any data analysis: input data is assembled, a computation is performed upon it, and the result is transmitted elsewhere. The elements do not interact in any unconventional manner, and the claim recites no technology-based solution of the kind found in BASCOM Global Internet Services, Inc. v. AT&T Mobility LLC, 827 F.3d 1341, 1350 (Fed. Cir. 2016), or DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1258-59 (Fed. Cir. 2014). Accordingly, claim 1 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 2
Step 1: Claim 2 depends from claim 1 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 2 contains no additional abstract idea limitations.
Step 2A, Prong 2: Claim 2 recites the following additional elements, “receive at least a portion of the interaction data from the computing system via the communications interface; and store the received portion of the interaction data within the memory.”
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 1, the additional elements carried over from claim 1 do not integrate the judicial exception into a practical application.
With respect to the new additional elements, “receive at least a portion of the interaction data from the computing system via the communications interface” is mere pre-solution data gathering: it obtains the very data upon which the abstract idea operates and imposes no meaningful limit on how the exception is applied. Such data gathering is insignificant extra-solution activity. See MPEP 2106.05(g). The limitation “store the received portion of the interaction data within the memory” recites a generic computer component performing its most basic and ordinary function. See MPEP 2106.05(f). Neither limitation, alone or in combination with the elements of claim 1, reflects any improvement to computer functionality or to any other technology; the specification attributes no technical effect to the receipt or storage of the interaction data. Accordingly, claim 2 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 1, the additional elements carried over from claim 1 are well-understood, routine, and conventional. With respect to “receive at least a portion of the interaction data from the computing system via the communications interface,” receiving data over a network has been recognized by the courts as well-understood, routine, and conventional. See Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1321 (Fed. Cir. 2016); TLI Communications LLC v. AV Automotive LLC, 823 F.3d 607, 610 (Fed. Cir. 2016); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355 (Fed. Cir. 2014); MPEP 2106.05(d)(II). With respect to “store the received portion of the interaction data within the memory,” storing and retrieving information in memory and electronic recordkeeping have likewise been recognized as well-understood, routine, and conventional. See Versata Development Group, Inc. v. SAP America, Inc., 793 F.3d 1306, 1334 (Fed. Cir. 2015); OIP Technologies, Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363 (Fed. Cir. 2015); Alice Corp., 573 U.S. at 225-26; MPEP 2106.05(d)(II). The specification confirms the same, stating that “[t]he essential elements of a computer are a central processing unit for performing or executing instructions and one or more memory devices for storing instructions and data” and that components “can be interconnected by any form or medium of digital data communication, such as a communication network,” including “a local area network (LAN) and a wide area network (WAN), such as the Internet” (¶[0158]; ¶[0161]). The ordered combination adds nothing further. Accordingly, claim 2 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 3
Step 1: Claim 3 depends from claim 1 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 3 additionally recites the following abstract idea limitations, “generate the input dataset in accordance with the data that characterizes the composition”. Assembling a set of values in a specified arrangement is an act of organizing information that a person can practically perform in the mind or with pen and paper, and is therefore a mental process of the same nature as the corresponding limitation of claim 1. See MPEP 2106.04(a)(2)(III).
Step 2A, Prong 2: Claim 3 recites the following additional elements, “obtain (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset” and “apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters”. The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 1, the additional elements carried over from claim 1 do not integrate the judicial exception into a practical application.
With respect to the new additional elements, “obtain (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset” is mere pre-solution data gathering — the retrieval of configuration values before the abstract idea is performed — and is insignificant extra-solution activity. See MPEP 2106.05(g). The limitation “apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters” does not narrow the artificial intelligence process in any technically meaningful way: the parameters are unspecified, no value or type of parameter is recited, and applying a process “in accordance with” its own parameters is inherent in applying it at all. The limitation therefore remains a generic instruction to apply the exception using a generic class of computer algorithm. See MPEP 2106.05(f). Under Ex Parte Desjardins, the specification was again consulted; it attributes no technical effect to the retrieval of process parameters or to the parameterized application of the process, and claim 3 accordingly reflects no disclosed improvement. See MPEP 2106.05(a). Accordingly, claim 3 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 1, the additional elements carried over from claim 1 are well-understood, routine, and conventional. With respect to “obtain (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset,” obtaining such configuration information amounts to storing and retrieving information in memory where it is retrieved locally, and to receiving data over a network where it is obtained from another system. Both constructions are addressed because the term “obtain” is limited to neither under the broadest reasonable interpretation, and the courts have recognized both activities as well-understood, routine, and conventional. See Versata Development Group, Inc. v. SAP America, Inc., 793 F.3d 1306, 1334 (Fed. Cir. 2015); OIP Technologies, Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363 (Fed. Cir. 2015); Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1321 (Fed. Cir. 2016); TLI Communications LLC v. AV Automotive LLC, 823 F.3d 607, 610 (Fed. Cir. 2016); MPEP 2106.05(d)(II). With respect to “apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters,” the performance of repetitive calculations by a computer is well-understood, routine, and conventional. See Parker v. Flook, 437 U.S. 584, 594 (1978); Bancorp Servs., L.L.C. v. Sun Life Assurance Co. of Canada (U.S.), 687 F.3d 1266, 1278 (Fed. Cir. 2012); MPEP 2106.05(d)(II). The specification confirms this, identifying the process as “a gradient-boosted decision-tree process (e.g., XGBoost process)” (¶[019]) whose “process coefficients, parameters, thresholds, and other process parameters” are generated and stored in a conventional data store (¶[049]). The ordered combination adds nothing further. Accordingly, claim 3 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 4
Step 1: Claim 4 depends from claim 3 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 4 incorporates all the limitations of claims 1 and 3 and is therefore directed to the same abstract idea identified in claim 1. Claim 4 additionally recites the following abstract idea limitations, “based on the data that characterizes the composition, perform operations that at least one of extract a first feature value from the interaction data or compute a second feature value based on the first feature value; and generate the input dataset based on at least one of the extracted first feature value or the computed second feature value.” These are further steps in the assembly of the input dataset. Reading a value out of a record, deriving a second value from that value, and placing the result into a dataset are observations, evaluations, and simple calculations that a person can practically perform in the mind or with pen and paper, and are therefore mental processes of the same nature as the corresponding limitations of claim 1. See MPEP 2106.04(a)(2)(III). To the extent the recited computation of “a second feature value based on the first feature value” is characterized as a calculation, it also falls within the mathematical concepts grouping. See MPEP 2106.04(a)(2)(I).
Step 2A, Prong 2 & Step 2B: No new additional elements are introduced. The analysis from the parent claim is maintained.
CLAIM 5
Step 1: Claim 5 depends from claim 1 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 5 incorporates all the limitations of claim 1 and is therefore directed to the same abstract idea identified in claim 1. Claim 5 additionally recites the following abstract idea limitations, “the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.” Unlike the generic recitation in claim 1, this limitation names a specific algorithm — gradient boosting over decision trees — and therefore sets forth and describes mathematical calculations by name. It accordingly recites an additional judicial exception falling within the mathematical concepts grouping. See MPEP 2106.04(a)(2)(I). This is the pattern of USPTO Subject Matter Eligibility Example 47, claim 2, in which training “wherein the selected training algorithm includes a backpropagation algorithm and a gradient descent algorithm” was found to recite a mathematical concept because the calculations were named, and it is distinguishable from Example 39, in which generic training language recited none. No new additional elements are introduced; the added limitation instead narrows what was treated in claim 1 as an additional element into the judicial exception itself. It is acknowledged that this characterization is not free from doubt. Example 47's qualifying language named two specific calculation procedures, whereas “gradient-boosted, decision-tree process” names a model family without setting forth a formula or equation in words or mathematical symbols, and a reasonable alternative reading would place it nearer Example 39. The rejection does not depend on the resolution of that question. If the limitation is instead treated as an additional element rather than as part of the judicial exception, it fails to integrate the exception into a practical application for the reasons given in the Step 2A, Prong 2 analysis of the parent claim, and it is well-understood, routine, and conventional for the reasons given in the Step 2B analysis below.
Step 2A, Prong 2: The judicial exception is not integrated into a practical application. Claim 5 does not introduce any new additional elements beyond those in claim 1, and it removes rather than adds material available for the integration inquiry, because the artificial intelligence process is now part of the recited exception. See MPEP 2106.04(d)(2). Naming the mathematical technique used does not confer a technological improvement: the specification identifies the gradient-boosted decision-tree process as a known technique, “an ensemble or decision-tree process, such as a gradient-boosted decision-tree process (e.g., XGBoost process)” (¶[019]), and describes no modification to how that process is structured or executed. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 1, the additional elements do not integrate the judicial exception into a practical application. Accordingly, claim 5 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. No new additional elements are introduced. For the same reasons as discussed in the Step 2B analysis of claim 1, the additional elements are well-understood, routine, and conventional. The recited gradient-boosted decision-tree process is not itself an additional element, and in any event the specification expressly identifies it as a known, commercially available technique (¶[019]), which is Berkheimer-compliant evidence of the first type. See Berkheimer v. HP Inc., 881 F.3d 1360 (Fed. Cir. 2018); MPEP 2106.05(d)(I). The ordered combination adds nothing further. Accordingly, claim 5 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 6
Step 1: Claim 6 depends from claim 1 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 6 incorporates all the limitations of claim 1 and is therefore directed to the same abstract idea identified in claim 1. Claim 6 additionally recites obtaining elements of second interaction data bearing temporal identifiers, partitioning those elements between a training interval and a validation interval, generating training datasets, obtaining targeting data, and training the artificial intelligence process. The limitations “based on the temporal identifiers, determine that a first subset of the elements of the second interaction data are associated with a prior training interval, and that a second subset of the elements of the second interaction data are associated with a prior validation interval” and “generate a plurality of training datasets based on corresponding portions of the first subset” recite the judicial exception: sorting dated records into two groups by inspecting their dates, and assembling the sorted records into sets, are evaluations and acts of organizing information that a person can practically perform in the mind or with pen and paper. See MPEP 2106.04(a)(2)(III).
Step 2A, Prong 2: Claim 6 recites the following additional elements, “obtain elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval”, “obtain elements of targeting data identifying each of the targeted events”, and “perform operations that train the artificial intelligence process based on the training datasets and the targeting data”.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claim 1, the additional elements carried over from claim 1 do not integrate the judicial exception into a practical application.
With respect to the new additional elements, “obtain elements of second interaction data … comprising a temporal identifier” and “obtain elements of targeting data identifying each of the targeted events” are mere pre-solution data gathering and the selection of a particular type of data to be manipulated, both of which are insignificant extra-solution activity. See MPEP 2106.05(g). With respect to “perform operations that train the artificial intelligence process based on the training datasets and the targeting data,” the claim recites the bare fact that training occurs. It identifies no objective function, no update rule, no architecture, and no measure of what a successful training run produces; it recites the idea of a trained model rather than a particular way of producing one. Such a limitation covers only the desired outcome, and does not improve technology or a technical field. See MPEP 2106.05(a); McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1314-15 (Fed. Cir. 2016).
Under Ex Parte Desjardins, the specification was again consulted with respect to these limitations. The specification describes the temporally partitioned training and validation of the process (¶[019]; ¶[020]), but the improvement it attributes to the disclosed embodiments remains the throughput of the parallelized, distributed GPU/TPU implementation described at ¶[022] and ¶¶[026]-[027] — none of which claim 6 recites. Unlike the claims in Ex Parte Desjardins, claim 6 recites no adjustment to model parameters, no protection of previously learned knowledge, no reduction in storage, and no reduction in system complexity; it recites only that some training occurs on data sorted by date. Claim 6 therefore does not reflect any disclosed technological improvement. See MPEP 2106.05(a); Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1316 (Fed. Cir. 2016). Accordingly, claim 6 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claim 1, the additional elements carried over from claim 1 are well-understood, routine, and conventional. With respect to the obtaining limitations, receiving or retrieving data over a network and from memory is well-understood, routine, and conventional. See Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1321 (Fed. Cir. 2016); TLI Communications LLC v. AV Automotive LLC, 823 F.3d 607, 610 (Fed. Cir. 2016); Versata Development Group, Inc. v. SAP America, Inc., 793 F.3d 1306, 1334 (Fed. Cir. 2015); MPEP 2106.05(d)(II). With respect to “perform operations that train the artificial intelligence process,” the performance of repetitive calculations by a computer is well-understood, routine, and conventional. See Parker v. Flook, 437 U.S. 584, 594 (1978); Bancorp Servs., L.L.C. v. Sun Life Assurance Co. of Canada (U.S.), 687 F.3d 1266, 1278 (Fed. Cir. 2012); MPEP 2106.05(d)(II). The specification independently confirms the conventional nature of the recited training, describing it as training of “a gradient-boosted decision-tree process (e.g., XGBoost process)” (¶[019]) performed on commercially available platforms, namely “an Apache Spark™ distributed, cluster-computing framework, a Databricks™ analytical platform” (¶[026]). The ordered combination adds nothing further. Accordingly, claim 6 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 7
Step 1: Claim 7 depends from claim 6 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 7 incorporates all the limitations of claims 1 and 6 and is therefore directed to the same abstract idea identified in claim 1. Claim 7 additionally recites the following abstract idea limitations, “generate a plurality of validation datasets based on portions of the second subset”, “compute one or more validation metrics based on the additional elements of output data”, and “based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process”, each of which recites the judicial exception. Assembling records into validation sets is an act of organizing information; computing a metric from a set of results and comparing that metric against a threshold to reach a pass/fail conclusion are calculations and evaluations that a person can practically perform in the mind or with pen and paper. These are mental processes, and the computation of a metric additionally falls within the mathematical concepts grouping. See MPEP 2106.04(a)(2)(I), (III).
Step 2A, Prong 2: : Claim 7 recites the following additional elements, “apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets”.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claims 1 and 6, the additional elements carried over from those claims do not integrate the judicial exception into a practical application.
With respect to the new additional element, “apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data” is the same generic application of a generic class of computer algorithm already analyzed in claim 1, differing only in the data to which the process is applied. Selecting a particular type of data to be manipulated does not integrate the exception, and applying a generic algorithm in its ordinary capacity amounts to the equivalent of an instruction to “apply it.” See MPEP 2106.05(f), (g). The specification attributes no technical effect to the validation step beyond confirming “a predictive capability, and an accuracy, of the adaptively trained … process” (¶[079]), which is the ordinary purpose of validating any model; claim 7 therefore reflects no disclosed technological improvement. See MPEP 2106.05(a). Accordingly, claim 7 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claims 1 and 6, the additional elements carried over from those claims are well-understood, routine, and conventional. With respect to “apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data,” the performance of repetitive calculations by a computer is well-understood, routine, and conventional. See Parker v. Flook, 437 U.S. 584, 594 (1978); Bancorp Servs., L.L.C. v. Sun Life Assurance Co. of Canada (U.S.), 687 F.3d 1266, 1278 (Fed. Cir. 2012); MPEP 2106.05(d)(II). The specification confirms the conventional nature of the recited validation, describing the computed metrics as ordinary recall- and precision-based values (¶[079]). The ordered combination adds nothing further. Accordingly, claim 7 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
CLAIM 8
Step 1: Claim 8 depends from claim 1 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 8 incorporates all the limitations of claim 1 and is therefore directed to the same abstract idea identified in claim 1. Claim 8 additionally recites the following abstract idea limitations, “the plurality of targeted events comprise a first targeted redemption event, a second targeted redemption event, and a third targeted redemption event, each of the first, second, and third targeted redemption events; and the output data comprises a first numerical score indicative of a predicted likelihood of an occurrence of the first targeted redemption event during the second temporal interval, a second numerical score indicative of a predicted likelihood of an occurrence of the second, and a third numerical score indicative of a predicted likelihood of an occurrence of the third targeted redemption event during the second temporal interval.” These limitations specify that the events forecast are redemptions of a financial product and that the forecast takes the form of three numerical likelihood scores. Redemption of a financial product is a fundamental economic practice and a commercial interaction, and assigning a numerical likelihood to each of three possible outcomes is an evaluation and judgment that a person can practically perform in the mind or with pen and paper. The added limitations therefore recite the same abstract idea, further specified. See MPEP 2106.04(a)(2)(II), (III).
Step 2A, Prong 2 & Step 2B: No new additional elements are introduced. The analysis from the parent claim is maintained.
CLAIM 9
Step 1: Claim 9 depends from claim 8 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 9 incorporates all the limitations of claims 1 and 8 and is therefore directed to the same abstract idea identified in claim 1. Claim 9 additionally recites the following abstract idea limitations, “each of the first, second, and third targeted redemption events is associated with a product; and the first targeted redemption event corresponds to a full redemption of the product during the second temporal interval, the second targeted redemption event corresponds to a partial redemption of the product during the second temporal interval, and the third targeted redemption event corresponds to a non-occurrence of the full or partial redemption of the product during the second temporal interval.” These limitations do no more than define the three forecast outcomes as full redemption, partial redemption, and no redemption of a product. Defining the outcomes of a commercial transaction in a financial product is a fundamental economic practice and a commercial interaction, and falls within the certain methods of organizing human activity grouping. See MPEP 2106.04(a)(2)(II). No new additional elements are introduced.
Step 2A, Prong 2 & Step 2B: No new additional elements are introduced. The analysis from the parent claim is maintained.
CLAIM 10
Step 1: Claim 10 depends from claim 1 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 10 incorporates all the limitations of claim 1 and is therefore directed to the same abstract idea identified in claim 1. Claim 10 additionally recites the following abstract idea limitations, “the input dataset comprises feature values associated with a plurality of input features; and the explainability data comprises a feature contribution value characterizing a contribution of each of the input feature values to the predicted likelihood of the occurrences of the targeted events during the second temporal interval.” These limitations specify that the compiled dataset consists of feature values and that the contribution of every such value to the forecast is characterized. Assessing how much each of several considered factors contributed to a conclusion is an evaluation and judgment that a person can practically perform in the mind or with pen and paper, and therefore recites the same mental process identified in claim 1, merely applied to every feature rather than at least one. See MPEP 2106.04(a)(2)(III). No new additional elements are introduced.
Step 2A, Prong 2 & Step 2B: No new additional elements are introduced. The analysis from the parent claim is maintained.
CLAIM 11
Step 1: Claim 11 depends from claim 10 and recites an apparatus, which falls within the statutory category of a machine. See MPEP 2106.03.
Step 2A, Prong 1: Claim 11 incorporates all the limitations of claims 1 and 10 and is therefore directed to the same abstract idea identified in claim 1. Claim 11 additionally recites the following abstract idea limitations, “each of the targeted events are associated with a redemption of a product”, which further specifies the abstract idea as the fundamental economic practice of redeeming a financial product (see MPEP 2106.04(a)(2)(II)); “select a subset of the features values based on an application of a factor analysis process to the feature contribution values”, which names a specific statistical technique — factor analysis — and therefore sets forth a mathematical concept, and which also recites the mental act of selecting a subset (see MPEP 2106.04(a)(2)(I), (III)); and “generate clustering data characterizing a plurality of customer clusters, each of the customer clusters being associated with a corresponding redemption profile”, which recites the mathematical concept of clustering and, in grouping customers into behavioral redemption profiles for commercial use, also falls within the certain methods of organizing human activity grouping (see MPEP 2106.04(a)(2)(I), (II)).
Step 2A, Prong 2: The claim recites the following additional elements, “transmit the input dataset to the computing system” and the “application of a trained machine-learning or artificial-intelligence process to the subset of the feature values and to corresponding elements of the output data”.
The judicial exception is not integrated into a practical application. For the same reasons as discussed in the Step 2A, Prong 2 analysis of claims 1 and 10, the additional elements carried over from those claims do not integrate the judicial exception into a practical application.
With respect to the new additional elements, “transmit the input dataset to the computing system” is post-solution transmission of data that has already been assembled, and is insignificant extra-solution activity. See MPEP 2106.05(g). The “application of a trained machine-learning or artificial-intelligence process to the subset of the feature values and to corresponding elements of the output data” is again the invocation of a generic class of computer algorithm in its ordinary capacity to carry out the recited grouping, with no architecture, training regime, or computation specified; it amounts to the equivalent of an instruction to “apply it.” See MPEP 2106.05(f). Under Ex Parte Desjardins, the specification was again consulted; it describes the clustering only as an application of a known technique, “a clustering algorithm, such as a k-means clustering algorithm” (¶[021]), used to establish customer “redemption personas” that “further facilitate a proactive engagement of customers of the financial institution” (¶[021]). That is a commercial benefit, not an improvement to computer functionality or to any other technology, and claim 11 accordingly reflects no disclosed technological improvement. See MPEP 2106.05(a). Accordingly, claim 11 does not satisfy Step 2A, Prong 2.
Step 2B: The claim does not include additional elements that amount to significantly more than the judicial exception. For the same reasons as discussed in the Step 2B analysis of claims 1 and 10, the additional elements carried over from those claims are well-understood, routine, and conventional. With respect to “transmit the input dataset to the computing system,” transmitting data over a network is well-understood, routine, and conventional. See Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307, 1321 (Fed. Cir. 2016); TLI Communications LLC v. AV Automotive LLC, 823 F.3d 607, 610 (Fed. Cir. 2016); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355 (Fed. Cir. 2014); MPEP 2106.05(d)(II). With respect to the “application of a trained machine-learning or artificial-intelligence process,” the performance of repetitive calculations by a computer is well-understood, routine, and conventional. See Parker v. Flook, 437 U.S. 584, 594 (1978); Bancorp Servs., L.L.C. v. Sun Life Assurance Co. of Canada (U.S.), 687 F.3d 1266, 1278 (Fed. Cir. 2012); MPEP 2106.05(d)(II). The specification confirms the conventional nature of the recited process by identifying it as “a clustering algorithm, such as a k-means clustering algorithm” (¶[021]), which is Berkheimer-compliant evidence of the first type. See Berkheimer v. HP Inc., 881 F.3d 1360 (Fed. Cir. 2018); MPEP 2106.05(d)(I). The ordered combination adds nothing further. Accordingly, claim 11 does not satisfy Step 2B and is rejected under 35 U.S.C. 101.
Claims 12, 13, 14, 15, 16, 17, 18 and 19 are substantially similar in scope and spirit to claims 1, 3, 4, 5, 6, 7, 8, 10+11, respectively. Therefore, the 35 U.S.C. 101 rejections of claims 1, 3, 4, 5, 6, 7, 8, 10+11 are substantively applied accordingly. The only difference in the rejections are in the Step 1 analysis, where claims 12-19 are directed to a “A computer-implemented method …”, followed by a series of steps, therefore falling within the statutory category of a process. See MPEP 2106.03.
Claim 20 is substantially similar in scope and spirit to claim 1. Therefore, the 35 U.S.C. 101 rejection of claim 1 is substantively applied accordingly. The only difference in the rejection is in the Step 1 analysis, where claim 20 is “A tangible, non-transitory computer-readable medium storing instructions …”, that is, an article of manufacture, therefore falls within the statutory category of a manufacture. See MPEP 2106.03.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-8 and 10-20 are rejected under 35 USC 103 as being unpatentable over US Pat. Pub. No. 2018/0046926A1 to Achin et al. (hereinafter Achin) in view of US Pat. Pub. No. 2011/0270779A1 to Showalter and further in view of Explainable AI for Trees: From Local Explanations to Global Understanding Lundberg et al. (hereinafter Lundberg).
Per claim 1, Achin discloses An apparatus (Achin: ¶[0042]…Achin discloses a predictive modeling apparatus having a memory and at least one processor that together carry out a predictive modeling procedure, which constitutes claimed the apparatus under BRI, "Other embodiments of this aspect include a predictive modeling apparatus including: a memory configured to store a machine-executable module encoding a predictive modeling procedure, wherein the predictive modeling procedure includes a plurality of tasks including at least one pre-processing task and at least one model-fitting task; and at least one processor configured to execute the machine-executable module, wherein executing the machine-executable module causes the apparatus to perform the predictive modeling procedure"), comprising:
a memory storing instructions (Achin: ¶[0042]…Achin's memory stores a machine-executable module encoding the predictive modeling procedure, and a PHOSITA would understand a machine-executable module to be stored program instructions, which constitutes the claimed memory storing instructions, "a memory configured to store a machine-executable module encoding a predictive modeling procedure, wherein the predictive modeling procedure includes a plurality of tasks including at least one pre-processing task and at least one model-fitting task");
a communications interface (Achin: ¶[0230]…Achin discloses an interface services layer through which external systems reach the prediction module, supply new observations to it and receive the returned predictions, which constitutes the claimed communications interface under BRI, "In some embodiments, users and external systems may access a prediction module (e.g., in an interface services layer of predictive modeling system 100), specify one or more predictive models to be used, and supply new observations. The prediction module may then return the predictions provided by those models."); and
at least one processor coupled to the memory and the communications interface, the at least one processor being configured to execute the instructions to (Achin: ¶[0042]…Achin's processor executes the machine-executable module held in the memory and thereby causes the apparatus to perform the recited operations, "at least one processor configured to execute the machine-executable module, wherein executing the machine-executable module causes the apparatus to perform the predictive modeling procedure"):
generate an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval (Achin: ¶[0042]…Achin assembles the model's input observations from time-series data in which every observation carries an indication of the time associated with it, so the observations drawn from a bounded input time range constitute the claimed input dataset generated from first data associated with a first temporal interval under BRI, " Performing the predictive modeling procedure may include performing the pre-processing task, including: (a) obtaining time-series data including one or more data sets, wherein each data set includes a plurality of observations, wherein each observation includes (1) an indication of a time associated with the observation and (2) respective values of one or more variables");
based on an application of the trained artificial intelligence process to the input dataset, generate output data (Achin: ¶[0342]…Achin applies the fitted predictive model over a user-specified forecast range and generates predicted values across that range, which constitutes generating output data from an application of the trained process to the input dataset, "The user may indicate a desired forecast range (e.g., the number of future time periods to be predicted by the model or the number of distinct future events to be predicted by the model)")…the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval (Achin: ¶[0367]…Achin's skip range is expressly a temporal lag interposed between the latest observation on which the predictions are based and the earliest prediction in the forecast range, and Achin elsewhere calls that same skip range a gap between the end of the input window and the start of the later window, which constitutes the claimed buffer interval, "a skip range associated with a prediction problem represented by the time-series data is determined. The skip range may indicate a temporal lag between a time associated with an earliest prediction in the forecast range and a time associated with a latest observation upon which predictions in the forecast range are to be based"; ¶[0341]…Achin restates the separation as an express gap and confirms prediction begins only after it elapses, "The user may indicate a “skip range” in the data, which is a gap between the end of a training window (e.g., a time range of data used for training) and the start of a validation window (e.g., a time range of data used for validation) or a holdout window (e.g., a time range of data used for holdout testing). In any case, the practical goal may be to begin predicting after a certain gap from the last available historical observation.");
transmit at least a portion of the output data (Achin: ¶[0230]…Achin returns the model's predictions through the prediction module to the external systems that requested them and describes communicating those predictions out to the operating locations, "In some embodiments, users and external systems may access a prediction module (e.g., in an interface services layer of predictive modeling system 100), specify one or more predictive models to be used, and supply new observations. The prediction module may then return the predictions provided by those models."; ¶[0341]…the predictions are communicated to the receiving locations, "But operationally, even after the predictive model is finished, it may take several days to both communicate predictions to all the store locations and for the stores to make adjustments to their operations in response to the predictions.")…and the computing system being configured to perform operations based on the portion of the output data (Achin: ¶[0341]…Achin's receiving store locations act on the transmitted predictions by making adjustments to their operations, "But operationally, even after the predictive model is finished, it may take several days to both communicate predictions to all the store locations and for the stores to make adjustments to their operations in response to the predictions.")…
Achin does not expressly disclose, but Showalter does teach:
…generate an output representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval (Showalter: ¶[0098]…Showalter's personalization models are expressly distinguished from a conventional single-target score on the ground that they predict a portfolio of target events at once, so the model output carries a prediction for each of a plurality of targeted events, which constitutes the claimed limitation under BRI, "Whereas a typical score predicts only one kind of target event (e.g. FICO, default), the personalization models simultaneously predict a portfolio of target events across a wide range of borrower, asset and local market behaviors"; ¶[0308]…each such prediction is expressly a likelihood, "Scores predict the likelihood of a target event. These probability calculations assume the event will occur and are based on machine learned matrix.; ¶[0309]…the likelihood is scoped to a defined future outcome period, which constitutes the claimed second temporal interval, "The borrower default score predicts first lien default behavior of a one year outcome period").
Achin combined with Showalter does not expressly disclose, but with Lundberg does teach:
…and explainability data associated with the trained artificial intelligence process to a computing system via the communications interface, the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data (Lundberg: Section 2.5, p. 5…Lundberg's TreeExplainer computes, for each individual prediction, an attribution value for every input feature of that prediction's input vector, and those attribution values are constrained to sum to the model output for that input, so they quantify the contribution of each input feature value to the specific output produced, "Local accuracy states that when approximating the original model f for a specific input x, the explanation’s attribution values should sum up to the output f(x)."; Section 1, p. 1…Lundberg frames these as local explanations of the impact of input features on individual predictions, "Yet while there is a rich history of global interpretation methods for trees that summarize the impact of input features on the model as a whole, much less attention has been paid to local explanations that explain the impact of input features on individual predictions (i.e. for a single sample)");
…and the computing system being configured to perform operations based on ... the explainability data (Lundberg: Section 2.7.5, p. 10…Lundberg feeds the per-prediction attribution values forward into downstream analytics that embed each sample into an explanation space and cluster it, which constitutes a receiving system performing operations on the basis of the explainability data, "We can address both of these limitations by using local explanation embeddings to embed each sample into a new “explanation space.” If we run clustering in this new space, we will get a supervised clustering where samples are grouped together based on their explanations").
Achin, Showalter and Lundberg are analogous art because they are each from the same field of endeavor, specifically the training and application of machine-learning models to historical time-stamped records in order to predict future outcomes and to act on those predictions. They are further reasonably pertinent to the same problem with which the inventor was involved, namely producing a forward-looking, actionable and interpretable prediction from a subject's accumulated historical data. Lundberg expressly situates the tree-based models it explains within this same field, including finance and customer retention (Lundberg: Section 1, p. 1…Lundberg identifies the deployment fields of the tree-based models to which its explanation method applies, "Random forests, gradient boosted trees, and other tree-based models are used in finance, medicine, biology, customer retention, advertising, supply chain management, manufacturing, public health, and many other areas to make predictions based on sets of input features.").
Before the effective filing date of the claimed invention, it would have been obvious to a person having ordinary skill in the art to have configured Achin's predictive modeling apparatus to emit, for each of a plurality of targeted events, the per-event likelihood score Showalter teaches. This is a combination of prior art elements according to known methods to yield the predictable result of a multi-event probabilistic forecast (KSR rationale A).
The suggestion/motivation for doing so would have been provided by the applied references themselves. Showalter expressly frames the portfolio-of-target-events output as an improvement over a conventional single-target score (Showalter: ¶[0098]…Showalter states the deficiency of single-target scoring that its multiple-target approach is designed to remedy, "Whereas a typical score predicts only one kind of target event (e.g. FICO, default), the personalization models simultaneously predict a portfolio of target events across a wide range of borrower, asset and local market behaviors"), and Achin's own architecture already contemplates fitting and serving multiple models against the same time-series dataset (Achin: ¶[0230]…Achin permits a requester to specify one or more predictive models to be used and returns the predictions of those models together, "In some embodiments, users and external systems may access a prediction module (e.g., in an interface services layer of predictive modeling system 100), specify one or more predictive models to be used, and supply new observations. The prediction module may then return the predictions provided by those models."). A PHOSITA serving predictions to an operating system that must act on them would therefore have adopted Showalter's per-event likelihood output within Achin's platform.
Before the effective filing date of the claimed invention, it would have been obvious to a person having ordinary skill in the art to have further configured the Achin/Showalter apparatus to compute and transmit, alongside the predictions, the per-feature attribution values taught by Lundberg. This is the use of a known technique to improve a similar system in the same way (KSR rationale C).
The suggestion/motivation for doing so would have been provided by the applied references themselves. Achin already computes feature-importance values for its fitted models but only at the level of the model as a whole (Achin: ¶[0183]…Achin measures the degree of significance each feature has in predicting the target, and does so at the level of the fitted model as a whole rather than for any individual prediction, "Variable importance, which measures the degree of significance each feature has in predicting the target, may be analyzed using “gradient boosted trees”, Breiman and Cutler’s “Random Forest’, “alternating conditional expectations”, and/or other suitable techniques."), and Lundberg identifies precisely this shortfall as the problem it solves and supplies the per-prediction remedy (Lundberg: Section 1, p. 1…Lundberg identifies the gap between global feature-importance summaries and local per-prediction explanations and offers the latter, "Yet while there is a rich history of global interpretation methods for trees that summarize the impact of input features on the model as a whole, much less attention has been paid to local explanations that explain the impact of input features on individual predictions (i.e. for a single sample)").
Per claim 2, Achin combined with Showalter and Lundberg discloses claim 1. Achin further teaches receive at least a portion of the interaction data from the computing system via the communications interface; and store the received portion of the interaction data within the memory (Achin: ¶[0230]…Achin’s external systems supply new observations inward through the prediction module, "In some embodiments, users and external systems may access a prediction module (e.g., in an interface services layer of predictive modeling system 100), specify one or more predictive models to be used, and supply new observations. The prediction module may then return the predictions provided by those models"; ¶[0320]…the datasets so received are committed to the apparatus’s own file-storage module, which Achin identifies as the store for uploaded datasets, "Types of data stored via this module include uploaded datasets, derived data, model computations, and predictions.").
Per claim 3, Achin combined with Showalter and Lundberg discloses claim 1. Achin further teaches obtain (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset; generate the input dataset in accordance with the data that characterizes the composition; and apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters (Achin: ¶[0231]…Achin retains, for each model, both the fitted coefficient and hyper-parameter values that characterize the trained model and the record of the modeling technique that dictates how each instance of new input data is assembled, and applies both when generating predictions on new observations, "For each model, exploration engine 110 may store a record of the modeling technique used to generate the model and the state of model the after fitting, including coefficient and hyper-parameter values"; [0015]…Achin separately fixes which of the available variables are admitted to the model as features and which are the targets, and that designation is the data that characterizes the composition of the dataset the model is then run on, "identifying one or more of the variables as targets, and identifying zero or more other variables as features").
Per claim 4, Achin combined with Showalter and Lundberg discloses claim 3. Achin further teaches based on the data that characterizes the composition, perform operations that at least one of extract a first feature value from the interaction data or compute a second feature value based on the first feature value; and generate the input dataset based on at least one of the extracted first feature value or the computed second feature value (Achin: ¶[0184]…Achin derives additional features from the dataset's existing variables by interpreting each variable's logical type and applying transformations to it, and admits the derived features into the dataset the model consumes, "Feature generation techniques may include generating additional features by interpreting the logical type of the dataset’s variable and applying various transformations to the variable").
Per claim 5, Achin combined with Showalter and Lundberg discloses claim 1. Achin further teaches the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process (Achin: ¶[0255]…Achin expressly names boosted trees among the model types its predictive modeling system fits and evaluates, so a gradient-boosted decision-tree process is among the trained processes Achin’s apparatus produces, "In some embodiments, the predictive modeling system 100 can fit many different model types, including, without limitation, decision trees, neural networks, support vector machine models, regression models, boosted trees, random forests, deep learning neural networks, etc."; ¶[0195]…Achin further tunes that model type by its own hyper-parameters, confirming the gradient boosted trees model is one the system actually fits rather than merely mentions, "Examples of hyper-parameters include, without limitation, the penalty parameters of an elastic-net model, the number of trees in a gradient boosted trees model, the number of neighbors in a nearest neighbors model, etc.").
Per claim 6, Achin combined with Showalter and Lundberg discloses claim 1. Achin further teaches obtain elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval; based on the temporal identifiers, determine that a first subset of the elements of the second interaction data are associated with a prior training interval, and that a second subset of the elements of the second interaction data are associated with a prior validation interval; generate a plurality of training datasets based on corresponding portions of the first subset; obtain elements of targeting data identifying each of the targeted events; and perform operations that train the artificial intelligence process based on the training datasets and the targeting data (¶[0015]…Achin separately and expressly designates which of the variables are the targets to be predicted, which constitutes the claimed obtaining of targeting data identifying each of the targeted events under BRI, "identifying one or more of the variables as targets, and identifying zero or more other variables as features").
Per claim 7, Achin combined with Showalter and Lundberg discloses claim 6. Achin further teaches generate a plurality of validation datasets based on portions of the second subset; apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets; compute one or more validation metrics based on the additional elements of output data; and based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process (¶[0256]…the ranked validation-set assessments are then applied against a stated acceptance criterion, namely best individual performance, and only the model meeting that criterion is selected, so the measured validation metric is tested for consistency with a threshold condition before the model is accepted, "The choice of the final model can be made by the predictive modeling system 100 or by the user";¶[0256]…the criterion itself, "the opportunity to build ensemble models from those component models that exhibit the best individual performance").
Per claim 8, Achin combined with Showalter and Lundberg discloses claim 1. Achin does not expressly disclose, but Showalter does teach:
the plurality of targeted events comprise a first targeted redemption event, a second targeted redemption event, and a third targeted redemption event (Showalter: ¶[0098] …Showalter's personalization models predict a portfolio of distinct target events for the same borrower and financial product, among them a default event, a prepayment event and a refinance event, which constitutes the recited first, second and third targeted events under BRI, "Whereas a typical score predicts only one kind of target event (e.g. FICO, default), the personalization models simultaneously predict a portfolio of target events across a wide range of borrower, asset and local market behaviors."; ¶[0309]…the first such event and its score, "The borrower default score predicts first lien default behavior of a one year outcome period."; ¶[0315]…a further distinct event and its score, "Another score using machine learned matrix is the borrower refinance score. The borrower refinance score predicts the likelihood of a refinance over a one year period."; ¶[0312]…a further such event is expressly the payoff of the borrower’s mortgage, i.e. the redemption of the financial product the borrower holds, which anchors the recited events as targeted redemption events rather than merely targeted events, "The borrower prepayment score predicts the likelihood that the borrower’s mortgage (regardless of whether current or in default) will be paid off during the next 12 months."); and the output data comprises a first numerical score indicative of a predicted likelihood of an occurrence of the first targeted redemption event during the second temporal interval, a second numerical score indicative of a predicted likelihood of an occurrence of the second, and a third numerical score indicative of a predicted likelihood of an occurrence of the third targeted redemption event during the second temporal interval (Showalter: ¶[0308]…Showalter emits a separate numerical score per target event, each expressly a likelihood of that event over a defined future outcome period, which constitutes the recited first, second and third numerical scores, "Scores predict the likelihood of a target event. These probability calculations assume the event will occur and are based on machine learned matrix"; ¶[0309]…the default score over a one year outcome period, "The borrower default score predicts first lien default behavior of a one year outcome period"; ¶[0315]…the refinance score over a one year period, "Another score using machine learned matrix is the borrower refinance score. The borrower refinance score predicts the likelihood of a refinance over a one year period."). The rationale to combine Showalter with Achin is the same as the parent claim.
Per claim 10, Achin combined with Showalter and Lundberg discloses claim 1. Showalter combined with Lundberg further teaches the input dataset comprises feature values associated with a plurality of input features; and the explainability data comprises a feature contribution value characterizing a contribution of each of the input feature values to the predicted likelihood of the occurrences of the targeted events during the second temporal interval. Under the combination, the model output to which those attribution values are constrained to sum is the per-event likelihood score Showalter supplies (Showalter: ¶[0308]…each such output is a likelihood of a target event, "Scores predict the likelihood of a target event. These probability calculations assume the event will occur and are based on machine learned matrix"), so each feature contribution value characterizes a contribution to the predicted likelihood of the occurrences of the targeted events during the second temporal interval as the limitation requires). The rationale to combine Showalter/Lundberg with Achin is the same as the parent claim.
Per claim 11, Achin combined with Showalter and Lundberg discloses claim 10, Achin further disclosing: the at least one processor is further configured to execute the instructions to transmit the input dataset to the computing system (Achin: ¶[0320]…Achin’s file-storage module holds the uploaded input dataset and the distributed cloud workers that operate on it obtain the dataset from that module into their own local storage, so the apparatus conveys the input dataset out to the computing components that consume it, which constitutes the claimed transmission of the input dataset to the computing system under BRI, "Types of data stored via this module include uploaded datasets, derived data, model computations, and predictions"; ¶[0320]…the conveyance to the consuming components is express, "when cloud workers access this module, they may also temporarily cache the stored files in their local storage").
Achin combined with Showalter does not expressly disclose, but Lundberg does teach:
select a subset of the features values based on an application of a factor analysis process to the feature contribution values (Lundberg: Section 2.7.5, p. 10…Lundberg runs principal component analysis over the local explanation embeddings, the per-feature contribution values, and thereby resolves them into a reduced set of underlying risk-factor categories, and principal component analysis is a factor-analysis process within the BRI of that term, "Analogously, we can also run PCA on local explanation embeddings for chronic kidney disease samples, which uncovers the two primary categories of risk factors that identify unique individuals at risk of end-stage renal disease"); and
based on an application of a trained machine-learning or artificial-intelligence process to the subset of the feature values and to corresponding elements of the output data, generate clustering data (Lundberg: Section 2.7.5, p. 10…Lundberg embeds each sample into an explanation space built from its contribution values and clusters the samples in that space, yielding groups of subjects that share both a common predicted outcome and a common explanation for it, "We can address both of these limitations by using local explanation embeddings to embed each sample into a new “explanation space.” If we run clustering in this new space, we will get a supervised clustering where samples are grouped together based on their explanations"; Section 2.7.5, p. 10…the resulting clusters are characterized by their shared outcome and shared reasons, "Running hierarchical supervised clustering using the mortality model results in many groups of people that share a similar mortality risk for similar reasons").
Achin and Lundberg does not expressly disclose, but Showalter does teach:
each of the targeted events are associated with a redemption of a product (Showalter: ¶[0312]…Showalter’s prepayment score is expressly a prediction that the borrower’s mortgage, which is the financial product the borrower holds, will be paid off within the scored period, and the payoff of a held financial product is a redemption of that product under BRI, "The borrower prepayment score predicts the likelihood that the borrower’s mortgage (regardless of whether current or in default) will be paid off during the next 12 months");
characterizing a plurality of customer clusters, each of the customer clusters being associated with a corresponding redemption profile (Showalter: ¶[0009]…Showalter’s cluster model sorts each borrower, who is a customer holding the lender’s financial product, into one of a plurality of borrower-property clusters, which constitutes the claimed plurality of customer clusters under BRI, "The cluster model includes an unsupervised machine-learned classifier configured to classifying a borrower of the loan and the specific property into one of a plurality of borrower-property clusters"; ¶[0309]…Showalter then pairs each such cluster group with its own predicted set of contract-disposition outcomes, so the cluster is not merely a partition but carries the outcome behaviour predicted for its members, "The score is used to predict the survival rate for each type of treatment (loan modification, short sale, third party sale, or foreclosure) paired with each borrower/property cluster group"; ¶[0313]…and the per-cluster redemption score is computed from the cluster assignment together with the factor values, so each cluster is associated with a corresponding profile of the redemption behaviour predicted for that cluster, which constitutes the claimed corresponding redemption profile under BRI, "The cluster assignment and factors (22) are used to calculate the borrower prepayment score.").
As it pertains to claim 11, before the effective filing date of the claimed invention, it would have been obvious to a PHOSITA to characterize the explanation-space clusters that Lundberg generates as the customer clusters, each carrying a corresponding redemption profile, that Showalter generates. This is the use of a known technique to improve a similar system in the same way (KSR rationale C).
The suggestion/motivation for doing so would have been provided by the applied references themselves. Lundberg teaches that clustering in the explanation space yields groups whose members share both a predicted outcome and the reasons for it, but leaves the population and the outcome to be supplied by the deploying application (Lundberg: Section 2.7.5, p. 10…Lundberg states the result of the clustering in terms of a shared risk and shared reasons without fixing the subject population, "Running hierarchical supervised clustering using the mortality model results in many groups of people that share a similar mortality risk for similar reasons."), and Showalter supplies exactly that specification for the financial-product context by sorting the customers of the product into clusters and pairing each cluster with its own predicted disposition outcomes (Showalter: ¶[0309]…each cluster group carries its own predicted outcome set, "The score is used to predict the survival rate for each type of treatment (loan modification, short sale, third party sale, or foreclosure) paired with each borrower/property cluster group."). A PHOSITA operating the combined Achin/Showalter/Lundberg platform on the customers of a financial product would therefore have arrived at customer clusters each associated with a corresponding redemption profile, with a reasonable expectation of success because Showalter already computes its per-cluster redemption scores from the same kind of factor-derived feature values that Lundberg's explanation embeddings supply (Showalter: ¶[0267]…the factor values, the cluster assignment and the per-event redemption scores are produced by one and the same analytics run, "Running the analytics creates the credit factors, product factors, property factors, metrics, a borrower as an asset flag, a property as an asset flag, a cluster assignment, a borrower prepayment score, a borrower default score, a borrower refinance score, a raw net present value calculation (NPV), and a refined NPV.").
Claims 12-19 are substantially similar in scope and spirit as claims 1 and 3-11, differing only in that claims 12-19 recite the corresponding computer-implemented method rather than the apparatus that performs it.
Claim 20 is substantially similar in scope and spirit as claim 1, differing only in that it recites a tangible, non-transitory computer-readable medium storing instructions that cause the recited method to be performed.
Therefore the rejections of claims 1 and 3-11 are applied accordingly.
Claim 9 is rejected under 35 USC 103 as being unpatentable over Achin in view Showalter and Lundberg, as applied in the rejection of claim 8 above, and further in view of Deferred Annuity Persistency to Drinkwater.
Achin combined with Showalter and Lundberg discloses claim 8.
Achin combined with Showalter and Lundberg does not expressly disclose, but Drinkwater does teach:
each of the first, second, and third targeted redemption events is associated with a product (Drinkwater: p. 10, Introduction; p. 11, Table 1…every contract-activity category Drinkwater measures attaches to one deferred annuity contract, a financial product held by a customer of a financial institution, and the categories are reported as mutually exclusive shares of that same contract population, which constitutes each of the recited first, second and third targeted redemption events being associated with a product under BRI, “The study was designed to analyze the surrender activity of deferred annuities by selected product, customer, and distribution characteristics”);
the first targeted redemption event corresponds to a full redemption of the product during the second temporal interval, the second targeted redemption event corresponds to a partial redemption of the product during the second temporal interval, and the third targeted redemption event corresponds to a non-occurrence of the full or partial redemption of the product during the second temporal interval (Drinkwater: p. 10, Introduction…Drinkwater expressly divides the disposition of a single contract into surrender of all of the cash value and withdrawal of only a portion of the cash value, which constitutes the recited full redemption of the product and partial redemption of the product under BRI, “The study examines contracts for which all of the cash value is surrendered (full surrenders, free looks, and internal replacements) and contracts for which only a portion of the cash value is withdrawn (partial withdrawals)”; p. 11, Summary of Results and Table 1…Drinkwater observes each contract over a one-year forward interval and assigns it to exactly one of full surrender, partial withdrawal, or no surrender activity, the three shares summing with the minor terminal categories to 100 percent of the contract population, which constitutes the recited third targeted redemption event corresponding to a non-occurrence of the full or partial redemption of the product during the second temporal interval under BRI, “Across all years, 1.5 million (5.5 percent) incurred a full surrender”; p. 11, Summary of Results…the partial and non-occurrence shares of that same population, “In addition, 3.9 million (14.0 percent) had part of the cash value withdrawn, 36,656 (0.1 percent) were annuitized, 286,376 (1.0 percent) were surrendered due to death or disability, and 22.0 million (79.4 percent) had no surrender activity (Table 1)”).
As it pertains to claim 9, Achin, Showalter, Lundberg and Drinkwater are analogous art because they are each from the same field of endeavor, specifically the statistical and machine-learned modeling of customer behavior with respect to financial products, and because Drinkwater is in any event reasonably pertinent to the financial problem being described, namely defining the mutually exclusive redemption outcomes of a financial product held by a customer that a predictive model is to score over a future interval.
Before the effective filing date of the claimed invention, it would have been obvious to a PHOSITA to specify the first, second and third targeted redemption events scored by the combined predictive-modeling system of Achin, Showalter and Lundberg as the full redemption, partial redemption, and non-occurrence of redemption outcomes that Drinkwater identifies for a financial product held by a customer.
The suggestion/motivation for doing so would have been provided by Drinkwater itself, which teaches that resolving the activity on a customer’s financial product into these outcome categories is precisely what permits that activity to be quantified against customer and product variables, “Another purpose of the analysis is to quantify, through modeling procedures, the relationship between surrender activity and variables associated with surrender. Quantifying this relationship is useful in product design, valuation, investment management, evaluation of market performance, and corporate planning.” (Drinkwater: p. 10, Introduction), and which itself fits a statistical model over those same variables to that outcome, “Logistic regression models (binary logit) were created to predict full surrenders in 2004, based on product, annuitant, and distribution channel characteristics for fixed retail, variable retail, and employer-sponsored markets.” (Drinkwater: p. 76, Appendix B). Substituting Drinkwater’s full/partial/no-redemption outcome partition for Showalter’s default/prepayment/refinance target events is a combination of prior art elements according to known methods to yield predictable results per MPEP 2143(A), and rests on an express teaching, suggestion or motivation in the prior art per MPEP 2143(G), the combined system’s multi-target scoring machinery being indifferent to which mutually exclusive outcomes of the product the three scores are assigned to.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claims 1-10, 12-18 and 20 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-4, 6-7, 9-10, 13-14, 16-17 and 20 of copending Pat. App. No. 17/714,288 (reference application) filed on 6/2/2026. Although the claims at issue are not identical, they are not patentably distinct from each other because the instant claims recite subject matter that is encompassed by, and constitutes an obvious variant of, the claims of the reference application. Reference claim 1 recites every element of instant claim 1 — a memory storing instructions, a communications interface, and at least one processor coupled thereto that generates an input dataset for a trained artificial intelligence process from elements of interaction data associated with a first temporal interval, applies that trained process to the input dataset to generate output data representative of a predicted likelihood of an occurrence of an event during a second temporal interval that is subsequent to the first temporal interval and separated from it by a corresponding buffer interval, and transmits at least a portion of that output data together with explainability data characterizing the contribution of a feature value of the input dataset to that output data to a computing system that is configured to perform operations based thereon. Instant claim 1 differs from reference claim 1 in only two respects. First, instant claim 1 omits the identifier-driven retrieval, consolidated-interaction-data, composition-data, distributed-parallel-processing and real-time limitations of reference claim 1; omission of limitations broadens the instant claim rather than rendering it patentably distinct, and a claim that is merely broader in scope than a reference claim is not patentably distinct from it. Second, instant claim 1 recites a predicted likelihood of an occurrence of “each of a plurality of targeted events” whereas reference claim 1 recites the predicted likelihood of an occurrence of a single attrition event. Generating the predicted likelihood for a plurality of targeted events using the same trained artificial intelligence process is a mere duplication of the recited predictive step yielding an entirely predictable result, and the reference claims themselves already contemplate generating a plurality of output-data predictions from a plurality of input datasets in parallel (reference claim 22) and training the artificial intelligence process on ground-truth data indicating both the occurrence and the non-occurrence of the targeted event (reference claim 27). A person of ordinary skill in the art would therefore have found the invention of instant claim 1 to be an obvious variation of the invention of reference claim 1 at the time the invention was made. The correspondence between the conflicting claims is set forth below.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Instant Claim 1 ↔ Copending App. No. 17/714,288 Claim 1
Instant Application Claims
Copending App. No. 17/714,288 Claims
An apparatus, comprising:
An apparatus, comprising:
a memory storing instructions;
a memory storing instructions;
a communications interface; and
a communications interface; and
at least one processor coupled to the memory and the communications interface, the at least one processor being configured to execute the instructions to:
at least one processor coupled to the memory and to the communications interface, the at least one processor being configured to execute the instructions to:
generate an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval;
receive, from a computing system via the communications interface, an identifier of a targeted participant in a service, and based on the received identifier, obtain, from the memory, elements of consolidated interaction data associated with a first temporal interval and with the targeted participant, the targeted participant being associated with a value of a parameter of the service that exceeds a threshold value; obtain, from the memory, composition data associated with an input dataset for a trained artificial intelligence process, the composition data specifying a sequential order of a plurality of input features of the input dataset; generate, in accordance with the composition data, feature values for the input features of the input dataset based on the elements of consolidated interaction data;
based on an application of the trained artificial intelligence process to the input dataset, generate output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
perform operations, in parallel across a plurality of distributed computing components interconnected across a computing network, and in real-time upon the receipt of the identifier, that apply the trained artificial intelligence process to the feature values of the input dataset, and based on an application of the trained artificial intelligence process to the feature values dataset, that generate output data representative of a predicted likelihood of an occurrence of an attrition event involving the targeted participant during a second temporal interval, ... and the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
transmit at least a portion of the output data and explainability data associated with the trained artificial intelligence process to a computing system via the communications interface, the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data, and the computing system being configured to perform operations based on the portion of the output data and the explainability data.
transmit, to the computing system via the communications interface, a notification comprising at least a portion of the generated output data and elements of explainability data associated with the trained artificial intelligence process, the elements of explainability data comprising a feature contribution value characterizing a contribution of a corresponding one of the feature values to the predicted likelihood of the occurrence of the attrition event during the second temporal interval, and the notification comprising information that causes the computing system to perform operations in accordance with the portion of the output data and with the feature contribution value, that reduce the predicted likelihood of the occurrence of the attrition event during the second temporal interval.
Instant Claim 2 ↔ Copending App. No. 17/714,288 Claim 2
The apparatus of claim 1, wherein the at least one processor is further configured to:
The apparatus of claim 1, wherein the at least one processor is further configured to execute the instructions to:
receive at least a portion of the interaction data from the computing system via the communications interface; and
receive elements of interaction data from at least one additional computing system via the communications interface; generate the elements of consolidated interaction data based on an application of one or more pre-processing operations to the elements of interaction data; and
store the received portion of the interaction data within the memory.
store the elements of consolidated interaction data within the memory, the elements of consolidated interaction data being associated with the identifier.
Instant Claim 3 ↔ Copending App. No. 17/714,288 Claim 3
The apparatus of claim 1, wherein the at least one processor is further configured to execute the instructions to:
The apparatus of claim 1, wherein: ... the at least one processor is further configured to execute the instructions to:
obtain (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset;
obtain one or more parameters that characterize the trained artificial intelligence process; and [claim 1] obtain, from the memory, composition data associated with an input dataset for a trained artificial intelligence process, the composition data specifying a sequential order of a plurality of input features of the input dataset;
generate the input dataset in accordance with the data that characterizes the composition; and
generate, in accordance with the composition data, feature values for the input features of the input dataset based on the elements of consolidated interaction data;
apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters.
perform the operations, in parallel across a plurality of distributed computing components, and in real-time upon the receipt of the identifier, that apply the trained, gradient-boosted, decision-tree process to the feature values in accordance with the one or more parameters.
Instant Claim 4 ↔ Copending App. No. 17/714,288 Claim 4
The apparatus of claim 3, wherein the at least one processor is further configured to execute the instructions to:
The apparatus of claim 1, wherein the at least one processor is further configured to execute the instructions to:
based on the data that characterizes the composition, perform operations that at least one of extract a first feature value from the interaction data or compute a second feature value based on the first feature value; and
based on the data that characterizes the composition, perform operations that at least one of extract a first one of the feature values from the elements of consolidated interaction data or compute a second one of the feature values based on the first feature value; and
generate the input dataset based on at least one of the extracted first feature value or the computed second feature value.
generate the input dataset based on at least one of the extracted first feature value or the computed second feature value.
Instant Claim 5 ↔ Copending App. No. 17/714,288 Claim 3
The apparatus of claim 1, wherein the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.
The apparatus of claim 1, wherein: the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process; and ...
Instant Claim 6 ↔ Copending App. No. 17/714,288 Claim 6
The apparatus of claim 1, wherein the at least one processor is further configured to execute the instructions to:
The apparatus of claim 1, wherein the at least one processor is further configured to execute the instructions to:
obtain elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval;
obtain elements of interaction data, each of the elements of interaction data comprising a temporal identifier associated with a temporal interval;
based on the temporal identifiers, determine that a first subset of the elements of the second interaction data are associated with a prior training interval, and that a second subset of the elements of the second interaction data are associated with a prior validation interval;
based on the temporal identifiers, determine that a first subset of the elements of the interaction data are associated with a prior training interval, and that a second subset of the elements of interaction data are associated with a prior validation interval;
generate a plurality of training datasets based on corresponding portions of the first subset;
generate a plurality of training datasets based on corresponding portions of the first subset; and
obtain elements of targeting data identifying each of the targeted events; and
[claim 27] ... using elements of ground-truth data associated with corresponding ones of the training datasets, ... each of elements of ground truth data indicating an occurrence or a non-occurrence of the attrition event involving the corresponding participant during a target temporal interval.
perform operations that train the artificial intelligence process based on the training datasets and the targeting data.
perform operations, in parallel across the plurality of distributed computing components, that train the artificial intelligence process based on the training datasets.
Instant Claim 7 ↔ Copending App. No. 17/714,288 Claim 7
The apparatus of claim 6, wherein the at least one processor is further configured to execute the instructions to:
The apparatus of claim 6, wherein the at least one processor is further configured to execute the instructions to:
generate a plurality of validation datasets based on portions of the second subset;
generate a plurality of validation datasets based on portions of the second subset;
apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
perform operations, in parallel across the plurality of distributed computing components, that apply the trained artificial intelligence process to the plurality of validation datasets, and that generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
compute one or more validation metrics based on the additional elements of output data; and
compute one or more validation metrics based on the additional elements of output data; and
based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process.
based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process.
Instant Claim 8 ↔ Copending App. No. 17/714,288 Claim 10
The apparatus of claim 1, wherein the at least one processor is further configured to execute instructions to:
The apparatus of claim 1, wherein ...
the plurality of targeted events comprise a first targeted redemption event, a second targeted redemption event, and a third targeted redemption event, each of the first, second, and third targeted redemption events; and
[claim 1] ... a predicted likelihood of an occurrence of an attrition event involving the targeted participant during a second temporal interval, the occurrence of the attrition event during the second temporal interval corresponding to a decrease in the parameter value by a threshold percentage during the second temporal interval ...
the output data comprises a first numerical score indicative of a predicted likelihood of an occurrence of the first targeted redemption event during the second temporal interval, a second numerical score indicative of a predicted likelihood of an occurrence of the second, and a third numerical score indicative of a predicted likelihood of an occurrence of the third targeted redemption event during the second temporal interval.
the output data comprise a numerical score indicative of the predicted likelihood of the occurrence of the attrition event during the second temporal interval.
Instant Claim 9 ↔ Copending App. No. 17/714,288 Claim 10
The apparatus of claim 8, wherein: each of the first, second, and third targeted redemption events is associated with a product; and
[claim 1] ... the targeted participant being associated with a value of a parameter of the service that exceeds a threshold value ...
the first targeted redemption event corresponds to a full redemption of the product during the second temporal interval, the second targeted redemption event corresponds to a partial redemption of the product during the second temporal interval, and the third targeted redemption event corresponds to a non-occurrence of the full or partial redemption of the product during the second temporal interval.
[claim 1] the occurrence of the attrition event during the second temporal interval corresponding to a decrease in the parameter value by a threshold percentage during the second temporal interval; [claim 27] each of elements of ground truth data indicating an occurrence or a non-occurrence of the attrition event involving the corresponding participant during a target temporal interval.
Instant Claim 10 ↔ Copending App. No. 17/714,288 Claim 9
The apparatus of claim 1, wherein: the input dataset comprises feature values associated with a plurality of input features; and
[claim 1] the composition data specifying a sequential order of a plurality of input features of the input dataset; generate, in accordance with the composition data, feature values for the input features of the input dataset ...
the explainability data comprises a feature contribution value characterizing a contribution of each of the input feature values to the predicted likelihood of the occurrences of the targeted events during the second temporal interval.
the explainability data comprising feature contribution values characterizing contributions of corresponding ones of the feature values to the predicted likelihood of the occurrence of the attrition event during the second temporal interval.
Instant Claim 12 ↔ Copending App. No. 17/714,288 Claim 13
A computer-implemented method, comprising:
A computer-implemented method, comprising:
generating, using at least one processor, an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval;
receiving, using at least one processor, an identifier of a targeted participant in a service from a computing system, and based on the received identifier, obtaining, from a data repository, elements of consolidated interaction data associated with a first temporal interval and with the targeted participant ...; obtaining, from the data repository, and the using at least one processor, composition data associated with an input dataset for a trained artificial intelligence process ...; generating, using the at least one processor, and in accordance with the composition data, feature values for the input features of the input dataset based on the elements of consolidated interaction data;
based on an application of the trained artificial intelligence process to the input dataset, generating, using the at least one processor, output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
using the at least one processor, performing operations, in parallel across a plurality of distributed computing components ..., that apply the trained artificial intelligence process to the feature values of the input dataset, and that, based on an application of the trained artificial intelligence process to the feature values, generate output data representative of a predicted likelihood of an occurrence of an attrition event involving the targeted participant during a second temporal interval, ... and the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
transmitting, using the at least one processor, at least a portion of the output data and elements of explainability data associated with the trained artificial intelligence process to a computing system, the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data, and the computing system being configured to perform operations based on the portion of the output data and the explainability data.
using the at least one processor, transmitting, to a computing system, a notification comprising at least a portion of the generated output data and elements of explainability data associated with the trained artificial intelligence process, the elements of explainability data comprising a feature contribution value characterizing a contribution of a corresponding one of the feature values to the predicted likelihood ..., and the notification comprising information that causes the computing system to perform operations in accordance with the portion of the output data and with the feature contribution value ...
Instant Claim 13 ↔ Copending App. No. 17/714,288 Claim 14
The computer-implemented method of claim 12, wherein: the computer-implemented method further comprises, using the at least one processor, obtaining (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset;
The computer-implemented method of claim 13, wherein: the computer-implemented method further comprises: using the at least one processor, obtaining one or more parameters that characterize the trained artificial intelligence process; and
generating the input dataset comprises generating the input dataset in accordance with the data that characterizes the composition; and
generating the input dataset comprises generating the input dataset based on at least one of the extracted first feature value or the computed second feature value;
the computer-implemented method further comprises applying, using the at least one processor, the trained artificial intelligence process to the input dataset in accordance with the one or more parameters.
the computer-implemented method further comprises performing operations, using the at least one processor, and in parallel across the plurality of distributed computing components, that apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters in real-time upon the receipt of the identifier; and
Instant Claim 14 ↔ Copending App. No. 17/714,288 Claim 14
The computer-implemented method of claim 13, wherein: the computer-implemented method further comprises, based on the data that characterizes the composition, performing operations, using the at least one processor, that at least one of extract a first feature value from the interaction data or compute a second feature value based on the first feature value; and
based on the data that characterizes the composition, performing operations, using the at least one processor, that at least one of extract a first feature value from the first interaction data or compute a second feature value based on the first feature value;
generating the input dataset comprises generating the input dataset based on at least one of the extracted first feature value or the computed second feature value.
generating the input dataset comprises generating the input dataset based on at least one of the extracted first feature value or the computed second feature value;
Instant Claim 15 ↔ Copending App. No. 17/714,288 Claim 14
The computer-implemented method of claim 12, wherein the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.
the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process, and the output data comprises a numerical score indicative of the predicted likelihood of an occurrence of the attrition event during the second temporal interval.
Instant Claim 16 ↔ Copending App. No. 17/714,288 Claim 16
The computer-implemented method of claim 12, further comprising:
The computer-implemented method of claim 13, further comprising:
obtaining, using the at least one processor, elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval;
obtaining, using the at least one processor, elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval;
based on the temporal identifiers, determining, using the at least one processor, that a first subset of the elements of the second interaction data are associated with a prior training interval, and that a second subset of the elements of the second interaction data are associated with a prior validation interval;
based on the temporal identifiers, determining, using the at least one processor, that a first subset of the elements of the second interaction data is associated with a prior training interval, that a second subset of the elements of the second interaction data is associated with a prior validation interval;
generating, using the at least one processor, a plurality of training datasets based on corresponding portions of the first subset;
generating, using the at least one processor, a plurality of training datasets based on corresponding portions of the first subset; and
obtaining, using the at least one processor, elements of targeting data identifying each of the targeted events; and
[claim 27] ... using elements of ground-truth data associated with corresponding ones of the training datasets, ... each of elements of ground truth data indicating an occurrence or a non-occurrence of the attrition event ...
performing operations, using the at least one processor, that train the artificial intelligence process based on the training datasets and the targeting data.
performing operations, using the at least one processor, and in parallel across the plurality of distributed computing components, that train the artificial intelligence process based on the training datasets.
Instant Claim 17 ↔ Copending App. No. 17/714,288 Claim 17
The computer-implemented method of claim 16, further comprising:
The computer-implemented method of claim 16, further comprising:
generating, using the at least one processor, a plurality of validation datasets based on portions of the second subset;
generating, using the at least one processor, a plurality of validation datasets based on portions of the second subset;
using the at least one processor, applying the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
performing operations, using the at least one processor, and in parallel across the plurality of distributed computing components, that apply the trained artificial intelligence process to the plurality of validation datasets, and that generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
computing, using the at least one processor, one or more validation metrics based on the additional elements of output data; and
computing, using the at least one processor, one or more validation metrics based on the additional elements of output data; and
based on a determined consistency between the one or more validation metrics and a threshold condition, validating the trained artificial intelligence process using the at least one processor.
based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process using the at least one processor.
Instant Claim 18 ↔ Copending App. No. 17/714,288 Claim 14
The computer-implemented method of claim 12, wherein: the plurality of targeted events comprise a first targeted redemption event, a second targeted redemption event, and a third targeted redemption event, each of the first, second, and third targeted redemption events; and
[claim 13] ... output data representative of a predicted likelihood of an occurrence of an attrition event involving the targeted participant during a second temporal interval ...
the output data comprises a first numerical score indicative of a predicted likelihood of an occurrence of the first targeted redemption event during the second temporal interval, a second numerical score indicative of a predicted likelihood of an occurrence of the second, and a third numerical score indicative of a predicted likelihood of an occurrence of the third targeted redemption event during the second temporal interval.
the output data comprises a numerical score indicative of the predicted likelihood of an occurrence of the attrition event during the second temporal interval.
Instant Claim 20 ↔ Copending App. No. 17/714,288 Claim 20
A tangible, non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform a method, comprising:
A tangible, non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform a method, comprising:
generating an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval;
receiving an identifier of a targeted participant in a service from a computing system, and based on the received identifier, obtaining, from a data repository, elements of consolidated interaction data associated with a first temporal interval and with the targeted participant ...; obtaining, from the data repository, composition data associated with an input dataset for a trained artificial intelligence process ...; generating, in accordance with the composition data, feature values for the input features of the input dataset based on the elements of consolidated interaction data;
based on an application of the trained artificial intelligence process to the input dataset, generating output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
performing operations, in parallel across a plurality of distributed computing components ..., that apply the trained artificial intelligence process to the feature values of the input dataset, and that, based on an application of the trained artificial intelligence process to the feature values, generate output data representative of a predicted likelihood of an occurrence of an attrition event involving the targeted participant during a second temporal interval, ... and the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
transmitting at least a portion of the output data and elements of explainability data associated with the trained artificial intelligence process to a computing system, the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data, and the computing system being configured to perform operations based on the portion of the output data and the explainability data.
transmitting, to a computing system, a notification comprising at least a portion of the generated output data and elements of explainability data associated with the trained artificial intelligence process, the elements of explainability data comprising a feature contribution value characterizing a contribution of a corresponding one of the feature values to the predicted likelihood ..., and the notification comprising information that causes the computing system to perform operations in accordance with the portion of the output data and with the feature contribution value ...
Claims 1-8, 10, 12-18 and 20 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-6, 10-11, 13, 15-16 and 20 of copending Application No. 17/180,745 (reference application) filed on 12/30/2024. Although the claims at issue are not identical, they are not patentably distinct from each other because the conflicting claims are directed to the same underlying subject matter and differ only in ways that would have been obvious to one of ordinary skill in the art. Instant claim 1 and reference claim 1 both recite an apparatus having a memory storing instructions, a communications interface, and at least one processor coupled to the memory and the communications interface that generates an input dataset for a trained artificial intelligence process from elements of first interaction data associated with a first temporal interval, applies the trained artificial intelligence process to that input dataset to generate output data representative of a predicted likelihood of an occurrence of an event during a second temporal interval that is subsequent to the first temporal interval and separated from it by a corresponding buffer interval, and transmits at least a portion of that output data to a computing system via the communications interface, the computing system being configured to perform operations based on the transmitted output data. The instant claim differs in two respects. First, instant claim 1 omits the identifier-receipt, change-in-composition detection, metric-versus-threshold inconsistency determination, and process-parameter modification limitations of reference claim 1. That omission only broadens the instant claim, and a claim that omits limitations recited in a reference claim is an obvious variation of that reference claim; the instant claim would improperly extend the right to exclude already sought in the reference application. Second, instant claim 1 recites a predicted likelihood for each of a plurality of targeted events rather than for an event, and requires that explainability data characterizing a contribution of at least one feature value of the input dataset to the output data be transmitted with the output data. As to the plurality of targeted events, reference claim 1 already recites determining the predicted likelihood of an occurrence of an event, and performing that same recited determination for each of several events of interest is a mere duplication of a recited step that yields only the predictable result of a separate likelihood per event. As to the explainability data, reference claim 1 already requires computing “a value of one or more metrics characterizing the application of the ... trained artificial intelligence process to each of the input datasets” and already requires transmitting, to the computing system and onward to the associated device, both the elements of output data and further data derived from that output; transmitting the value that characterizes how the trained process acted on a given input dataset alongside the output data it explains, so that the computing system may act on both, would have been an obvious variation of what reference claim 1 already recites. Instant independent claims 12 and 20 recite the same subject matter as instant claim 1 in computer-implemented method and non-transitory computer-readable medium form and correspond to reference claims 13 and 20 in the same manner. The instant dependent claims correspond limitation-for-limitation to the reference claims identified in the table below.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Instant Claim 1 ↔ Copending App. No. 17/180,745 Claim 1
Instant Application Claims
Copending App. No. 17/180,745 Claims
An apparatus, comprising: a memory storing instructions; a communications interface; and at least one processor coupled to the memory and the communications interface, the at least one processor being configured to execute the instructions to:
An apparatus, comprising: a memory storing instructions; a communications interface; and at least one processor coupled to the memory and the communications interface, the at least one processor being configured to execute the instructions to:
generate an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval;
obtain, from the memory, elements of first interaction data associated with the identifiers and with one or more temporal intervals ...; based on the detected change in the composition, generate, for each of the identifiers, an input dataset based on corresponding ones of the elements of first interaction data associated with the first temporal interval;
based on an application of the trained artificial intelligence process to the input dataset, generate output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
apply a trained artificial intelligence process to each of the input datasets in accordance with a value of at least one process parameter, and based on the application of the trained artificial intelligence process to each of the input datasets, generate a corresponding element of output data representative of a predicted likelihood of an occurrence of an event during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval;
transmit at least a portion of the output data and explainability data associated with the trained artificial intelligence process to a computing system via the communications interface, ... and the computing system being configured to perform operations based on the portion of the output data and the explainability data.
transmit at least a subset of the elements of output data and corresponding ones of the identifiers across the communications network to the computing system via the communications interface, the computing system being configured to: at least one of modify an element of second interaction data associated with at least one of the identifiers based on the corresponding element of output data, or generate an additional element of the second interaction data ...;
... the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data ...
based on at least the elements of output data, compute a value of one or more metrics characterizing the application of the application of the trained artificial intelligence process to each of the input datasets; and
[Instant claim 1 omits the following limitations of reference claim 1; the omission broadens the instant claim.]
receive a plurality of identifiers from a computing system across a communications network via the communications interface ...; detect a change in a composition of the elements of first interaction data associated with at least a first one of the temporal intervals; ... transmit data characterizing the at least one of the modification or the generation ...; determine an inconsistency between the one or more metric values and at least one threshold condition, and perform operations that modify the at least one process parameter value in accordance with the determined inconsistency.
Instant Claim 2 ↔ Copending App. No. 17/180,745 Claim 2
The apparatus of claim 1, wherein the at least one processor is further configured to:
The apparatus of claim 1, wherein the at least one processor is further configured to:
receive at least a portion of the interaction data from the computing system via the communications interface; and
receive at least a portion of the first interaction data from the computing system via the communications interface; and
store the received portion of the interaction data within the memory.
store the received portion of the first interaction data within the memory.
Instant Claim 3 ↔ Copending App. No. 17/180,745 Claim 3
obtain (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset;
obtain (i) a value of one or more process parameters that characterize the trained artificial intelligence process and (ii) composition data that characterizes a composition of the input dataset of the trained artificial intelligence process;
generate the input dataset in accordance with the data that characterizes the composition; and
generate each of the input datasets in accordance with the composition data; and
apply the trained artificial intelligence process to the input dataset in accordance with the one or more parameters.
apply the trained artificial intelligence process to each of the input datasets in accordance with the one or more process parameter values.
Instant Claim 4 ↔ Copending App. No. 17/180,745 Claim 4
based on the data that characterizes the composition, perform operations that at least one of extract a first feature value from the interaction data or compute a second feature value based on the first feature value; and
based on the composition data, perform operations that at least one of extract a first feature value from the first interaction data or compute a second feature value based on the first feature value; and
generate the input dataset based on at least one of the extracted first feature value or the computed second feature value.
generate a corresponding one of the input datasets based on at least one of the extracted first feature value or the computed second feature value.
Instant Claim 5 ↔ Copending App. No. 17/180,745 Claim 6
The apparatus of claim 1, wherein the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.
The apparatus of claim 1, wherein the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.
Instant Claim 6 ↔ Copending App. No. 17/180,745 Claim 10
obtain elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval;
obtain elements of third interaction data, each of the elements of the third interaction data comprising a temporal identifier associated with a temporal interval;
based on the temporal identifiers, determine that a first subset of the elements of the second interaction data are associated with a prior training interval, and that a second subset of the elements of the second interaction data are associated with a prior validation interval;
based on the temporal identifiers, determine that a first subset of the elements of the third interaction data are associated with a prior training interval, and that a second subset of the elements of the third interaction data are associated with a prior validation interval; and
generate a plurality of training datasets based on corresponding portions of the first subset;
generate a plurality of training datasets based corresponding portions of the first subset, ...
obtain elements of targeting data identifying each of the targeted events; and
[No counterpart in the reference claims.]
perform operations that train the artificial intelligence process based on the training datasets and the targeting data.
... and perform operations that train the artificial intelligence process based on the training datasets.
Instant Claim 7 ↔ Copending App. No. 17/180,745 Claim 11
generate a plurality of validation datasets based on portions of the second subset;
generate a plurality of the validation datasets based on portions of the second subset;
apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
compute one or more validation metrics based on the additional elements of output data; and
compute one or more validation metrics based on the additional elements of output data; and
based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process.
based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process.
Instant Claim 8 ↔ Copending App. No. 17/180,745 Claim 5
the plurality of targeted events comprise a first targeted redemption event, a second targeted redemption event, and a third targeted redemption event, each of the first, second, and third targeted redemption events; and
[Reference claim 1 recites the predicted likelihood of an occurrence of “an event”; reciting three such events is a duplication of that recited determination.]
the output data comprises a first numerical score indicative of a predicted likelihood of an occurrence of the first targeted redemption event during the second temporal interval, a second numerical score ..., and a third numerical score indicative of a predicted likelihood of an occurrence of the third targeted redemption event during the second temporal interval.
The apparatus of claim 1, wherein the corresponding element of output data comprises a numerical score indicative of the predicted likelihood of the occurrence of the event during the second temporal interval.
Instant Claim 10 ↔ Copending App. No. 17/180,745 Claim 4
the input dataset comprises feature values associated with a plurality of input features; and
(claim 4) ... perform operations that at least one of extract a first feature value from the first interaction data or compute a second feature value based on the first feature value; and generate a corresponding one of the input datasets based on at least one of the extracted first feature value or the computed second feature value.
the explainability data comprises a feature contribution value characterizing a contribution of each of the input feature values to the predicted likelihood of the occurrences of the targeted events during the second temporal interval.
(claim 1) based on at least the elements of output data, compute a value of one or more metrics characterizing the application of the application of the trained artificial intelligence process to each of the input datasets; and
Instant Claim 12 ↔ Copending App. No. 17/180,745 Claim 13
A computer-implemented method, comprising: generating, using at least one processor, an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval;
A computer-implemented method, comprising: ... obtaining, using the at least one processor, elements of first interaction data associated with the identifiers and with one or more temporal intervals from a data repository, ...; based on the detected change in the composition, generating, for each of the identifiers, and using the at least one processor, an input dataset based on corresponding ones of the elements of first interaction data associated with the first temporal interval;
based on an application of the trained artificial intelligence process to the input dataset, generating, using the at least one processor, output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
using the at least one processor, applying a trained artificial intelligence process to each of the input datasets in accordance with a value of at least one process parameter, and ... generating, in real-time, a corresponding element of output data representative of a predicted likelihood of an occurrence of an event during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval;
transmitting, using the at least one processor, at least a portion of the output data and elements of explainability data associated with the trained artificial intelligence process to a computing system, ... and the computing system being configured to perform operations based on the portion of the output data and the explainability data.
transmitting, using the at least one processor, at least a subset of the elements of output data and corresponding ones of the identifiers across the communications network to the computing system, the computing system being configured to: at least one of modify an element of second interaction data ... or generate an additional element of the second interaction data ...;
... the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data ...
based on at least the elements of output data, computing, using the at least one processor, a value of one or more metrics characterizing the application of the application of the trained artificial intelligence process to each of the input datasets; and
Instant Claim 13 ↔ Copending App. No. 17/180,745 Claim 3
the computer-implemented method further comprises, using the at least one processor, obtaining (i) one or more parameters that characterize the trained artificial intelligence process and (ii) data that characterizes a composition of the input dataset;
obtain (i) a value of one or more process parameters that characterize the trained artificial intelligence process and (ii) composition data that characterizes a composition of the input dataset of the trained artificial intelligence process;
generating the input dataset comprises generating the input dataset in accordance with the data that characterizes the composition; and
generate each of the input datasets in accordance with the composition data; and
the computer-implemented method further comprises applying, using the at least one processor, the trained artificial intelligence process to the input dataset in accordance with the one or more parameters.
apply the trained artificial intelligence process to each of the input datasets in accordance with the one or more process parameter values.
Instant Claim 14 ↔ Copending App. No. 17/180,745 Claim 4
the computer-implemented method further comprises, based on the data that characterizes the composition, performing operations, using the at least one processor, that at least one of extract a first feature value from the interaction data or compute a second feature value based on the first feature value; and
based on the composition data, perform operations that at least one of extract a first feature value from the first interaction data or compute a second feature value based on the first feature value; and
generating the input dataset comprises generating the input dataset based on at least one of the extracted first feature value or the computed second feature value.
generate a corresponding one of the input datasets based on at least one of the extracted first feature value or the computed second feature value.
Instant Claim 15 ↔ Copending App. No. 17/180,745 Claim 15
The computer-implemented method of claim 12, wherein the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.
The computer-implemented method of claim 13, wherein the trained artificial intelligence process comprises a trained, gradient-boosted, decision-tree process.
Instant Claim 16 ↔ Copending App. No. 17/180,745 Claim 10
obtaining, using the at least one processor, elements of second interaction data, each of the elements of the second interaction data comprising a temporal identifier associated with a temporal interval;
obtain elements of third interaction data, each of the elements of the third interaction data comprising a temporal identifier associated with a temporal interval;
based on the temporal identifiers, determining, using the at least one processor, that a first subset ... are associated with a prior training interval, and that a second subset ... are associated with a prior validation interval;
based on the temporal identifiers, determine that a first subset of the elements of the third interaction data are associated with a prior training interval, and that a second subset ... are associated with a prior validation interval; and
generating, using the at least one processor, a plurality of training datasets based on corresponding portions of the first subset;
generate a plurality of training datasets based corresponding portions of the first subset, ...
obtaining, using the at least one processor, elements of targeting data identifying each of the targeted events; and
[No counterpart in the reference claims.]
performing operations, using the at least one processor, that train the artificial intelligence process based on the training datasets and the targeting data.
... and perform operations that train the artificial intelligence process based on the training datasets.
Instant Claim 17 ↔ Copending App. No. 17/180,745 Claim 11
generating, using the at least one processor, a plurality of validation datasets based on portions of the second subset;
generate a plurality of the validation datasets based on portions of the second subset;
using the at least one processor, applying the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
apply the trained artificial intelligence process to the plurality of validation datasets, and generate additional elements of output data based on the application of the trained artificial intelligence process to the plurality of validation datasets;
computing, using the at least one processor, one or more validation metrics based on the additional elements of output data; and
compute one or more validation metrics based on the additional elements of output data; and
based on a determined consistency between the one or more validation metrics and a threshold condition, validating the trained artificial intelligence process using the at least one processor.
based on a determined consistency between the one or more validation metrics and a threshold condition, validate the trained artificial intelligence process.
Instant Claim 18 ↔ Copending App. No. 17/180,745 Claim 16
the plurality of targeted events comprise a first targeted redemption event, a second targeted redemption event, and a third targeted redemption event, each of the first, second, and third targeted redemption events; and
[Reference claim 13 recites the predicted likelihood of an occurrence of “an event”; reciting three such events is a duplication of that recited determination.]
the output data comprises a first numerical score indicative of a predicted likelihood of an occurrence of the first targeted redemption event during the second temporal interval, a second numerical score ..., and a third numerical score indicative of a predicted likelihood of an occurrence of the third targeted redemption event during the second temporal interval.
each of the elements of output data comprises a numerical score indicative of the predicted likelihood of the occurrence of the event during the second temporal interval;
Instant Claim 20 ↔ Copending App. No. 17/180,745 Claim 20
A tangible, non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform a method, comprising:
A tangible, non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform a method, comprising:
generating an input dataset for a trained artificial intelligence process based on elements of first interaction data associated with a first temporal interval;
obtaining elements of first interaction data associated with the identifiers and with one or more temporal intervals from a data repository, ...; based on the detected change in the composition, generating for each of the identifiers, an input dataset based on corresponding ones of the elements of first interaction data associated with the first temporal interval;
based on an application of the trained artificial intelligence process to the input dataset, generating output data representative of a predicted likelihood of an occurrence of each of a plurality of targeted events during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval; and
applying a trained artificial intelligence process to each of the input datasets in accordance with a value of at least one process parameter, and ..., generating a corresponding element of output data representative of a predicted likelihood of an occurrence of an event during a second temporal interval, the second temporal interval being subsequent to the first temporal interval and being separated from the first temporal interval by a corresponding buffer interval;
transmitting at least a portion of the output data and elements of explainability data associated with the trained artificial intelligence process to a computing system, the explainability data characterizing a contribution of at least one feature value of the input dataset to the portion of the output data, and the computing system being configured to perform operations based on the portion of the output data and the explainability data.
transmitting at least a subset of the elements of output data and corresponding ones of the identifiers across the communications network to the computing system, ...; based on at least the elements of output data, computing a value of one or more metrics characterizing the application of the application of the trained artificial intelligence process to each of the input datasets; and
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALAN CHEN whose telephone number is (571)272-4143. The examiner can normally be reached M-F 10-7.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kamran Afshar can be reached at (571) 272-7796. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ALAN CHEN/ Primary Examiner, Art Unit 2125
Prosecution Insights