DETAILED ACTION
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 .
Applicant filed a response dated March 2, 2026 in which claims 1-20 have been amended. Therefore, claims 1-20 are currently pending in the application.
Priority
Application 18/748,727 was filed on June 20, 2024.
Examiner Request
The Applicant is requested to indicate where in the specification there is support for amendments to claims should Applicant amend. The purpose of this is to reduce potential 35 U.S.C. § 112(a) or § 112 1st paragraph issues that can arise when claims are amended without support in the specification. The Examiner thanks the Applicant in advance.
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 a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more. (MPEP 2106). The claims are directed to a method, system, and apparatus which is one of the statutory categories of invention (Step 1: YES). The recitation of the claimed invention is analyzed as follows, in which the abstract elements are boldfaced.
Claim 1 recites the limitations of:
A computer-implemented method comprising: accessing, by at least one of one or more servers of an electronic transaction system, electronic transaction data for a plurality of previous electronic transactions, the electronic transaction data comprising, for each previous electronic transaction, data for a plurality of transaction stages including at least a pre-authorization transaction stage and a subsequent transaction stage;
processing, by at least one of the one or more servers of the electronic transaction system, the electronic transaction data using a stage-differentiation model to generate stage-differentiated training data comprising, for each previous electronic transaction, data portions of the plurality of transaction stages;
associating, by at least one of the one or more servers of the electronic transaction system, each data portion for each previous electronic transaction with: (1) an action taken during a corresponding stage for the data portion, (2) an indication of the previous electronic transaction as fraudulent or non-fraudulent, and (3) a reward value determined as a function of the action taken, the corresponding stage, and the indication of the previous electronic transaction as fraudulent or non-fraudulent;
training, by at least one of the one or more servers of the electronic transaction system, a machine learning model using the stage-differentiated training data by updating the machine learning model based on the reward values associated with the actions taken during the stages of the previous electronic transactions; and
employing, by at least one of the one or more servers of the electronic transaction system, the machine learning model to determine values of impropriety for electronic transactions and allowing or blocking each electronic transaction based on the values of impropriety.
Claim 9 recites the limitations of:
A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising: receiving an indication of a current electronic transaction;
based on the indication, providing electronic transaction data, corresponding to the current electronic transaction, to a neural network for determining a value of impropriety for the current electronic transaction, the neural network having been trained using stage-differentiated training data generated from previous electronic transaction data for previous electronic transactions, the previous electronic transaction data for each previous electronic transaction comprising data for a plurality of transaction stages including at least a pre-authorization transaction stage and a subsequent transaction stage;
determining the value of impropriety for the current electronic transaction is above a threshold; and
blocking the current electronic transaction based on the value of impropriety being above a threshold.
Claim 15 recites the limitations of:
One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising: accessing, by at least one of one or more servers of an electronic transaction system, electronic transaction data for a plurality of previous electronic transactions, the electronic transaction data comprising, for each previous electronic transaction, data for a plurality of transaction stages including at least a pre-authorization transaction stage and a subsequent transaction stage;
processing, by at least one of the one or more servers of the electronic transaction system, the electronic transaction data using a stage-differentiation model to generate stage-differentiated training data comprising, for each previous electronic transaction, data portions of the plurality of transaction stages;
associating, by at least one of the one or more servers of the electronic transaction system, each data portion for each previous electronic transaction with: (1) an action taken during a corresponding stage for the data portion, (2) an indication of the previous electronic transaction as fraudulent or non-fraudulent, and (3) a reward value determined as a function of the action taken, the corresponding stage, and the indication of the previous electronic transaction as fraudulent or non-fraudulent;
training, by at least one of the one or more servers of the electronic transaction system, a machine learning model using the stage-differentiated training data by updating the machine learning model based on the reward values associated with the actions taken during the stages of the previous electronic transactions;
receiving, by at least one of the one or more servers of the electronic transaction system, current electronic transaction data for a current stage of a current electronic transaction;
providing, by at least one of the one or more servers of the electronic transaction system, the current electronic transaction data to the machine learning model to determine a value of impropriety for the current electronic transaction, and
causing, by at least one of the one or more servers of the electronic transaction system, a risk-assessment action to block or allow the current electronic transaction based on the value of impropriety.
The claim as a whole recites a method that, under its broadest reasonable interpretation, covers collecting, analyzing, and transmitting data to determine whether to block or allow a transaction, e.g., associated with a payment transaction. The specification discloses “[0002] At a high level, aspects described herein relate to systems, methods, and computer storage media for, among other things, determining whether an electronic transmission (e.g., associated with an electronic payment transaction) should be blocked (e.g., based on being a fraudulent electronic payment transaction)” and “[0025] In some embodiments, the electronic payment transaction client 102 can be associated with one or more of a seller interface and buyer interface (e.g., associated with an e-commerce platform).” This is a fundamental economic practice of a financial transaction; a commercial interaction, such as for business relations; and managing personal behavior or relationships or interactions between people, which are certain methods of organizing human activity.
Additionally, the claims recite the use of a trained a stage-differentiation model and machine-learning model for determining a value of impropriety for a transaction, e.g., a payment transaction. Claims 3, 11, and 16 also recite the use of Markov chain modeling to distinguish payment transaction data among stages. These are mathematical concepts or calculations.
In the alternative, the stage-differentiation model, machine learning model, and Markov chain model are considered a technology that is recited at a high level of generality and merely applied as a tool to implement the abstract idea.
Thus, the claims recite an abstract idea. (Step 2A, prong 1: YES).
Moreover, the judicial exception is not integrated into a practical application. Other than reciting a “A computer-implemented method comprising:”, “at least one of one or more servers of an electronic transaction system”, “A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising:”, “neural network”, and “One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising:” to perform the steps of “accessing”, “processing”, “associating”, “training”, “employing”, “determining”, “blocking”, and “providing”, nothing in the claim elements preclude the steps from practically being a certain method for organizing human activity, mental process, or mathematical calculation. The claim as a whole does not integrate the judicial exception into a practical application. The claim merely describes how to generally “apply” the concept of collecting, analyzing, and transmitting data to determine whether to block or allow a transmission, e.g., associated with a payment transaction in a computer environment. The additional computer elements recited in the claim limitations are recited at a high-level of generality such that it amounts to no more than mere instructions to apply the exception utilizing generic computer components.
For example, the Specification discloses “[0024] The electronic payment transaction client 102 may be a device that has the capability of accessing the network 108, and may also be referred to as a “computing device,” “mobile device,” “client device,” “user equipment (UE),” “communication device,” etc. The electronic payment transaction client 102 may, in some embodiments, take on a variety of forms, such as a personal computer, a laptop computer, a tablet, a mobile phone, a personal digital assistant, a server, or any other type of device that is capable of communication (e.g., by transmitting or receiving a signal) using the network 108. Broadly, the electronic payment transaction client 102 can include computer-readable media storing computer-executable instructions executed by at least one computer processor. One example of the electronic payment transaction client 102 includes computing device 500 described herein with reference to FIG. 5. The electronic payment transaction client 102 may be operated by a user, such as one or more of a person, machine, robot, another user device operator, or one or more combinations thereof.”
Furthermore, the Specification discloses “[0034] The policy-based reinforcement learning risk decision agent 110 can access the database 120 to execute tasks associated with one or more neural networks (e.g., reinforcement machine learning model(s) 122). For example, a user - via the electronic payment transaction client 102 (e.g., a prompt interface associated with the electronic payment interface 102A) - can communicate a request (e.g., a request to purchase a merchant offer on an e-commerce market) to the policy-based reinforcement learning risk decision agent 110 for processing of the request. Based on communicating the request, the policy-based reinforcement learning risk decision agent 110 can execute operations (e.g., via the risk decision generator 112, the electronic transaction blocker 114, or the electronic transaction facilitator 116) using one or more components of the database 120 (e.g., the reinforcement machine learning model(s) 122, historical electronic transaction data 124, or Markov chain data 132) - to facilitate or block one or more electronic transmissions associated with the request.”
Thus, the specification supports that general purpose computers or computer components are utilized to implement the steps of the abstract idea.
Merely implementing the abstract idea on a generic computer is not a practical application of the abstract idea. The claim as a whole, in viewing the additional elements both individually and in combination, does not integrate the judicial exception into a practical application. Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea. (Step 2A prong two: No)
The claim does not include additional elements, when considered both individually and as an ordered combination, that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of using “A computer-implemented method comprising:”, “at least one of one or more servers of an electronic transaction system”, “A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising:”, “neural network”, and “One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising:” to perform the steps of “accessing”, “processing”, “associating”, “training”, “employing”, “determining”, “blocking”, and “providing”, amounts to no more than mere instructions to apply the exception using generic computer component. The claim merely describes how to generally “apply” the concept of collecting, analyzing, and transmitting data to determine whether to block or allow a transmission, e.g., associated with a payment transaction in a computer environment. Thus, even when viewed as a whole, nothing in the claim adds significantly more (i.e. an inventive concept) to the abstract idea. Such additional elements are determined to not contain an inventive concept according to MPEP 2106.05(f). It should be noted that (1) the “recitation of claim limitations that attempt to cover any solution to an identified problem with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result, does not provide significantly more because this type of recitation is equivalent to the words “apply it”, and (2) “Use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice, commercial interaction, or managing personal behavior or relationships or interactions between people, mental process, or mathematical calculation) does not integrate a judicial exception into a practical application or provide significantly more”.
Dependent claims 2-8, 10-14, and 16-20 merely limit the abstract idea and do not recite any further additional elements beyond the cited abstract idea and the elements addressed above, thus, they do not amount to significantly more. The dependent claims are abstract for the reasons presented above because there are no additional elements that integrate the abstract idea into a practical application or are sufficient to amount to significantly more than the judicial exception when considered both individually and as an ordered combination. Thus, the dependent claims are directed to an abstract idea. (Step 2B: No)
Therefore, claims 1-20 are not patent-eligible.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. §§ 102 and 103 (or as subject to pre-AIA 35 U.S.C. §§ 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. § 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1-3, 5, 7, 9-10, 15, and 18 are rejected under 35 U.S.C. § 102(a)(1) as being anticipated by Vashisht, U.S. Patent Application Publication Number 2022/0261875.
As per claim 1,
Vashisht explicitly teaches:
A computer-implemented method comprising: accessing, by at least one of one or more servers of an electronic transaction system, electronic transaction data for a plurality of previous electronic transactions, the electronic transaction data comprising, for each previous electronic transaction, data for a plurality of transaction stages including at least a pre-authorization transaction stage and a subsequent transaction stage;
(Vashisht US20220261875 at paras. 72-75, 105-106) ("[0072] In one embodiment, the data pre-processing engine 220 is configured to randomly select all past transaction-level data (i.e., payment authorization request and payment authorization response messages of past payment transactions) associated with the issuer 108 and/or the merchant 104, for training the RL agent 222. The past transaction-level data associated with the issuer 108 and/or the merchant 104 is stored in the transaction database 118. In other words, the data pre-processing engine 220 is configured to access historical transactions and authorizing components (i.e., products) which were applied by the issuer 108 or the merchant 104 while processing a particular payment transaction. The transaction-level data associated with the issuer 108 or the merchant 104 includes a number of declined/approved/fraud transactions such that the RL agent 222 learns the apt representation of the transaction-level data associated with the issuer 108 or the merchant 104. [0073] In one embodiment, the data pre-processing engine 220 is configured to filter the past transaction-level data with some data constraints (such as, transaction type: card-not-present, decline reason code: addressable declines, issuer and/or merchant geographical region: for example, USA)." "[0105] At 522, the server system 200 accesses historical transaction data associated with the issuer 108 and/or merchant 104. The historical transaction data includes transaction-level data of past payment transaction requests and authorizing component information which were applied to each payment transaction request of the past payment transaction requests. As mentioned above, one or more authorizing components are configured to reduce the decline rates of payment transactions. [0106] At 524, the server system 200 aggregates the past payment transaction requests according to specific data elements. The specific data elements include, but are not limited to, such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector etc. The product flag vector for a particular payment transaction indicates authorizing components (i.e., products) that were enabled on the applied to the particular payment transaction.")
processing, by at least one of the one or more servers of the electronic transaction system, the electronic transaction data using a stage-differentiation model to generate stage-differentiated training data comprising, for each previous electronic transaction, data portions of the plurality of transaction stages;
(Vashisht US20220261875 at paras. 72-75, 83-96, 105-106) ("[0084] During the training phase (see, 314), the processor 206 is configured to access historical transaction data (see, 302) associated with the issuer 108 and/or the merchant 104 from the transaction database 118. The historical transaction data include, but is not limited to, past payment transactions (including authorization requests and authorization response details of the number of past payment transactions) of the issuer 108 for a particular time duration. In particular, the processor 206 is configured to extract various data elements (i.e., features) present in each payment transaction from the transaction-level data and perform data sanitization process (see, 306). The various data elements may include, but not limited to, transaction identifier, issuer name/identifier, merchant name/identifier, acquirer name/identifier, cross-border transaction flag (e.g., cross border, domestic), transaction channel flag (e.g., e-commerce, POS, recurring payments), payment card type (e.g., credit, debit), card product type (customer/commercial), card-not-present (CNP) transaction flag, response code flag (approve/decline), decline reason code (in case of declined transaction), etc. The various data elements are called as payment transaction attributes. Further, the historical transaction data also includes a product vector associated with each payment transaction that indicates what authorizing components (i.e., products) were applied to each payment transaction by the issuer 108 while performing the payment transaction. [0085] Further, the processor 206 is configured to filter-out transaction-level data of the past payment transactions for aggregation that have card-not-present (CNP) transaction flag indicating card-not-present transaction (see, filtration 308). The processor 206 is further configured to filter-out the transaction-level data of the past payment transactions which have a decline reason code indicating addressable declines (in a scenario, the payment transaction was declined). Thereafter, the processor 206 is configured to aggregate the filtered transaction-level data of the past payment transactions (see, 310) and provide the payment transaction attributes 312 and a product flag vector associated with each of the filtered past payment transactions to the deep reinforcement learning model for training.")
associating, by at least one of the one or more servers of the electronic transaction system, each data portion for each previous electronic transaction with: (1) an action taken during a corresponding stage for the data portion, (2) an indication of the previous electronic transaction as fraudulent or non-fraudulent, and (3) a reward value determined as a function of the action taken, the corresponding stage, and the indication of the previous electronic transaction as fraudulent or non-fraudulent;
(Vashisht US20220261875 at paras. 110-120) ("[0110] At 528b, the server system 200 defines an action space of the deep reinforcement learning model. The action space includes a plurality of actions. Each action corresponds to applying an authorizing component to a payment transaction. It should be noted that the action space corresponding to the action ‘a’ is not the application of the all authorizing components to the payment transaction available at the issuer 108." "[0116] As shown in Eqn. (2), the first term of the reward function includes p(Approval) and p(fraud). The second term of the reward function is inversely proportional to a summation of total cost of all authorizing components that may be applied to the payment transaction." "[0117] In one embodiment, the p(Approval) and p(fraud) are determined based on the historical transaction data of the issuer 108. The p(Approval) denotes a likelihood of getting a payment transaction approved after applying a particular product by the issuer 108 (see, table 3). For determining the approval and fraud probability scores, the server system 200 is configured to analyze past payment transactions and determine the number of approved transactions and declined transactions due to fraud for each payment transaction type, from the past payment transactions. In one example as shown in the table 3, an approval probability for a payment transaction type (cross-border, credit card, merchant industry) without applying any product is 0.4. In another example, an approval probability for the payment transaction type (cross border, credit card, merchant industry) after applying a product (e.g., “3D Secure, 3DS”) is 0.7. Similarly, fraud probability for a payment transaction type is also determined using existing fraud risk models." )
training, by at least one of the one or more servers of the electronic transaction system, a machine learning model using the stage-differentiated training data by updating the machine learning model based on the reward values associated with the actions taken during the stages of the previous electronic transactions; and
(Vashisht US20220261875 at paras. 104-110) ("[0104] FIG. 5B represents a flow chart 520 for training the deep reinforcement learning model 500, in accordance with an embodiment of the present disclosure. The sequence of operations of the flow chart 520 may not be necessarily executed in the same order as they are presented. Further, one or more operations may be grouped and performed in the form of a single step, or one operation may have several sub-steps that may be performed in parallel or in a sequential manner. [0105] At 522, the server system 200 accesses historical transaction data associated with the issuer 108 and/or merchant 104. The historical transaction data includes transaction-level data of past payment transaction requests and authorizing component information which were applied to each payment transaction request of the past payment transaction requests. As mentioned above, one or more authorizing components are configured to reduce the decline rates of payment transactions. [0106] At 524, the server system 200 aggregates the past payment transaction requests according to specific data elements. The specific data elements include, but are not limited to, such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector etc. The product flag vector for a particular payment transaction indicates authorizing components (i.e., products) that were enabled on the applied to the particular payment transaction. [0107] At 526, the server system 200 obtains payment transaction attributes based on the aggregated past payment transaction requests. The payment transaction attributes include, but are not limited to, information such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector, etc. [0108] At 528, the server system 200 trains the deep reinforcement learning model based, at least, on the payment transaction attributes and authorizing components (i.e., products) applied to the past payment transaction requests by the issuer 108. The training of the deep reinforcement learning model is performed at steps 528a-528d.")
employing, by at least one of the one or more servers of the electronic transaction system, the machine learning model to determine values of impropriety for electronic transactions and allowing or blocking each electronic transaction based on the values of impropriety.
(Vashisht US20220261875 at paras. 30-40, 117-118) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction.")
As per claim 2,
Vashisht explicitly teaches:
wherein a first electronic transaction is blocked during the pre-authorization transaction stage for the first electronic transaction based on a first value of impropriety determined for the first electronic transaction by the machine learning model.
(Vashisht US20220261875 at paras. 30-40, 99-102) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0101] In an example as shown in first row, the payment transaction features for a first payment transaction are cross-border, debit card, industry. An authorizing component is applied over the first payment transaction “Product_4” at a time, therefore, a product flag vector of the first payment transaction is 0001. Thus, payment transaction features and product information define a current state of the first payment transaction. The current state will get changed when the issuer 108 applies another authorizing component to the first payment transaction. In another example in the second row, the payment transaction features for a second payment transaction are domestic, credit card, industry. The product flag vector for the second payment transaction is 0010. Further, the industry code vector is a vector representation, where each index value refers to a particular industry type. In the first row, the industry code vector is 01000.")
As per claim 3,
Vashisht explicitly teaches:
wherein using the stage-differentiation model to generate the stage-differentiated training data comprises applying Markov chain modeling to each of the previous electronic transactions.
(Vashisht US20220261875 at paras. 77-79) ("[0078] In order to express the use of reinforcement learning in the product recommendation system for enhancing approval rates of the payment transaction more clearly, the present disclosure explains theoretical models of deep reinforcement learning model, the Markov Decision Process (MDP) with reference to FIG. 4 in more detail. It would be apparent to those skilled in the art that several of deep reinforcement learning models may be applied to accomplish the spirit of the present disclosure." "[0087] FIG. 4 is a block diagram representation of a deep reinforcement learning model 400, in accordance with an embodiment of the present disclosure. As shown in the FIG. 4, the deep reinforcement learning model involves two entities, i.e., an agent 402 (similar to the RL agent 222) and an environment 404, that interacts with each other. The agent 402 is an entity that makes product recommendation decisions, and the environment 404 may be set to feedback a reward value depending upon approval probability and fraud probability scores of a particular transaction and a cost associated with applying a combination of products to the particular transaction. The deep reinforcement learning model 400 implements Markov Decision Process (MDP).")
As per claim 5,
Vashisht explicitly teaches:
wherein employing the machine learning model to determine values of impropriety for the electronic transactions and allowing or blocking each electronic transaction based on the values of impropriety comprises: determining a first value of impropriety for a first electronic transaction is below a threshold; and facilitating the first electronic transaction based on the first value of impropriety being below the threshold.
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
As per claim 7,
Vashisht explicitly teaches:
wherein employing the machine learning model to determine values of impropriety for the electronic transactions and allowing or blocking each electronic transaction based on the values of impropriety comprises: determining a first value of impropriety for a first electronic transaction is above a threshold; and blocking the first electronic transaction based on the first value of impropriety being above the threshold.
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
As per claim 9,
Vashisht explicitly teaches:
A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising: receiving an indication of a current electronic transaction;
(Vashisht US20220261875 at paras. 30-40, 99-102) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0101] In an example as shown in first row, the payment transaction features for a first payment transaction are cross-border, debit card, industry. An authorizing component is applied over the first payment transaction “Product_4” at a time, therefore, a product flag vector of the first payment transaction is 0001. Thus, payment transaction features and product information define a current state of the first payment transaction. The current state will get changed when the issuer 108 applies another authorizing component to the first payment transaction. In another example in the second row, the payment transaction features for a second payment transaction are domestic, credit card, industry. The product flag vector for the second payment transaction is 0010. Further, the industry code vector is a vector representation, where each index value refers to a particular industry type. In the first row, the industry code vector is 01000.")
based on the indication, providing electronic transaction data, corresponding to the current electronic transaction, to a neural network for determining a value of impropriety for the current electronic transaction, the neural network having been trained using stage-differentiated training data generated from previous electronic transaction data for previous electronic transactions, the previous electronic transaction data for each previous electronic transaction comprising data for a plurality of transaction stages including at least a pre-authorization transaction stage and a subsequent transaction stage;
(Vashisht US20220261875 at paras. 72-75, 105-106) ("[0072] In one embodiment, the data pre-processing engine 220 is configured to randomly select all past transaction-level data (i.e., payment authorization request and payment authorization response messages of past payment transactions) associated with the issuer 108 and/or the merchant 104, for training the RL agent 222. The past transaction-level data associated with the issuer 108 and/or the merchant 104 is stored in the transaction database 118. In other words, the data pre-processing engine 220 is configured to access historical transactions and authorizing components (i.e., products) which were applied by the issuer 108 or the merchant 104 while processing a particular payment transaction. The transaction-level data associated with the issuer 108 or the merchant 104 includes a number of declined/approved/fraud transactions such that the RL agent 222 learns the apt representation of the transaction-level data associated with the issuer 108 or the merchant 104. [0073] In one embodiment, the data pre-processing engine 220 is configured to filter the past transaction-level data with some data constraints (such as, transaction type: card-not-present, decline reason code: addressable declines, issuer and/or merchant geographical region: for example, USA)." "[0105] At 522, the server system 200 accesses historical transaction data associated with the issuer 108 and/or merchant 104. The historical transaction data includes transaction-level data of past payment transaction requests and authorizing component information which were applied to each payment transaction request of the past payment transaction requests. As mentioned above, one or more authorizing components are configured to reduce the decline rates of payment transactions. [0106] At 524, the server system 200 aggregates the past payment transaction requests according to specific data elements. The specific data elements include, but are not limited to, such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector etc. The product flag vector for a particular payment transaction indicates authorizing components (i.e., products) that were enabled on the applied to the particular payment transaction.")
determining the value of impropriety for the current electronic transaction is above a threshold; and
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
blocking the current electronic transaction based on the value of impropriety being above a threshold.
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
As per claim 10,
Vashisht explicitly teaches:
wherein the current electronic transaction is blocked during a pre-authorization transaction stage of the current electronic transaction.
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
As per claim 15,
Vashisht explicitly teaches:
One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising: accessing, by at least one of one or more servers of an electronic transaction system, electronic transaction data for a plurality of previous electronic transactions, the electronic transaction data comprising, for each previous electronic transaction, data for a plurality of transaction stages including at least a pre-authorization transaction stage and a subsequent transaction stage;
(Vashisht US20220261875 at paras. 72-75, 105-106) ("[0072] In one embodiment, the data pre-processing engine 220 is configured to randomly select all past transaction-level data (i.e., payment authorization request and payment authorization response messages of past payment transactions) associated with the issuer 108 and/or the merchant 104, for training the RL agent 222. The past transaction-level data associated with the issuer 108 and/or the merchant 104 is stored in the transaction database 118. In other words, the data pre-processing engine 220 is configured to access historical transactions and authorizing components (i.e., products) which were applied by the issuer 108 or the merchant 104 while processing a particular payment transaction. The transaction-level data associated with the issuer 108 or the merchant 104 includes a number of declined/approved/fraud transactions such that the RL agent 222 learns the apt representation of the transaction-level data associated with the issuer 108 or the merchant 104. [0073] In one embodiment, the data pre-processing engine 220 is configured to filter the past transaction-level data with some data constraints (such as, transaction type: card-not-present, decline reason code: addressable declines, issuer and/or merchant geographical region: for example, USA)." "[0105] At 522, the server system 200 accesses historical transaction data associated with the issuer 108 and/or merchant 104. The historical transaction data includes transaction-level data of past payment transaction requests and authorizing component information which were applied to each payment transaction request of the past payment transaction requests. As mentioned above, one or more authorizing components are configured to reduce the decline rates of payment transactions. [0106] At 524, the server system 200 aggregates the past payment transaction requests according to specific data elements. The specific data elements include, but are not limited to, such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector etc. The product flag vector for a particular payment transaction indicates authorizing components (i.e., products) that were enabled on the applied to the particular payment transaction.")
processing, by at least one of the one or more servers of the electronic transaction system, the electronic transaction data using a stage-differentiation model to generate stage-differentiated training data comprising, for each previous electronic transaction, data portions of the plurality of transaction stages;
(Vashisht US20220261875 at paras. 72-75, 83-96, 105-106) ("[0084] During the training phase (see, 314), the processor 206 is configured to access historical transaction data (see, 302) associated with the issuer 108 and/or the merchant 104 from the transaction database 118. The historical transaction data include, but is not limited to, past payment transactions (including authorization requests and authorization response details of the number of past payment transactions) of the issuer 108 for a particular time duration. In particular, the processor 206 is configured to extract various data elements (i.e., features) present in each payment transaction from the transaction-level data and perform data sanitization process (see, 306). The various data elements may include, but not limited to, transaction identifier, issuer name/identifier, merchant name/identifier, acquirer name/identifier, cross-border transaction flag (e.g., cross border, domestic), transaction channel flag (e.g., e-commerce, POS, recurring payments), payment card type (e.g., credit, debit), card product type (customer/commercial), card-not-present (CNP) transaction flag, response code flag (approve/decline), decline reason code (in case of declined transaction), etc. The various data elements are called as payment transaction attributes. Further, the historical transaction data also includes a product vector associated with each payment transaction that indicates what authorizing components (i.e., products) were applied to each payment transaction by the issuer 108 while performing the payment transaction. [0085] Further, the processor 206 is configured to filter-out transaction-level data of the past payment transactions for aggregation that have card-not-present (CNP) transaction flag indicating card-not-present transaction (see, filtration 308). The processor 206 is further configured to filter-out the transaction-level data of the past payment transactions which have a decline reason code indicating addressable declines (in a scenario, the payment transaction was declined). Thereafter, the processor 206 is configured to aggregate the filtered transaction-level data of the past payment transactions (see, 310) and provide the payment transaction attributes 312 and a product flag vector associated with each of the filtered past payment transactions to the deep reinforcement learning model for training.")
associating, by at least one of the one or more servers of the electronic transaction system, each data portion for each previous electronic transaction with: (1) an action taken during a corresponding stage for the data portion, (2) an indication of the previous electronic transaction as fraudulent or non-fraudulent, and (3) a reward value determined as a function of the action taken, the corresponding stage, and the indication of the previous electronic transaction as fraudulent or non-fraudulent;
(Vashisht US20220261875 at paras. 110-120) ("[0110] At 528b, the server system 200 defines an action space of the deep reinforcement learning model. The action space includes a plurality of actions. Each action corresponds to applying an authorizing component to a payment transaction. It should be noted that the action space corresponding to the action ‘a’ is not the application of the all authorizing components to the payment transaction available at the issuer 108." "[0116] As shown in Eqn. (2), the first term of the reward function includes p(Approval) and p(fraud). The second term of the reward function is inversely proportional to a summation of total cost of all authorizing components that may be applied to the payment transaction." "[0117] In one embodiment, the p(Approval) and p(fraud) are determined based on the historical transaction data of the issuer 108. The p(Approval) denotes a likelihood of getting a payment transaction approved after applying a particular product by the issuer 108 (see, table 3). For determining the approval and fraud probability scores, the server system 200 is configured to analyze past payment transactions and determine the number of approved transactions and declined transactions due to fraud for each payment transaction type, from the past payment transactions. In one example as shown in the table 3, an approval probability for a payment transaction type (cross-border, credit card, merchant industry) without applying any product is 0.4. In another example, an approval probability for the payment transaction type (cross border, credit card, merchant industry) after applying a product (e.g., “3D Secure, 3DS”) is 0.7. Similarly, fraud probability for a payment transaction type is also determined using existing fraud risk models." )
training, by at least one of the one or more servers of the electronic transaction system, a machine learning model using the stage-differentiated training data by updating the machine learning model based on the reward values associated with the actions taken during the stages of the previous electronic transactions;
(Vashisht US20220261875 at paras. 104-110) ("[0104] FIG. 5B represents a flow chart 520 for training the deep reinforcement learning model 500, in accordance with an embodiment of the present disclosure. The sequence of operations of the flow chart 520 may not be necessarily executed in the same order as they are presented. Further, one or more operations may be grouped and performed in the form of a single step, or one operation may have several sub-steps that may be performed in parallel or in a sequential manner. [0105] At 522, the server system 200 accesses historical transaction data associated with the issuer 108 and/or merchant 104. The historical transaction data includes transaction-level data of past payment transaction requests and authorizing component information which were applied to each payment transaction request of the past payment transaction requests. As mentioned above, one or more authorizing components are configured to reduce the decline rates of payment transactions. [0106] At 524, the server system 200 aggregates the past payment transaction requests according to specific data elements. The specific data elements include, but are not limited to, such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector etc. The product flag vector for a particular payment transaction indicates authorizing components (i.e., products) that were enabled on the applied to the particular payment transaction. [0107] At 526, the server system 200 obtains payment transaction attributes based on the aggregated past payment transaction requests. The payment transaction attributes include, but are not limited to, information such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector, etc. [0108] At 528, the server system 200 trains the deep reinforcement learning model based, at least, on the payment transaction attributes and authorizing components (i.e., products) applied to the past payment transaction requests by the issuer 108. The training of the deep reinforcement learning model is performed at steps 528a-528d.")
receiving, by at least one of the one or more servers of the electronic transaction system, current electronic transaction data for a current stage of a current electronic transaction;
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
providing, by at least one of the one or more servers of the electronic transaction system, the current electronic transaction data to the machine learning model to determine a value of impropriety for the current electronic transaction, and
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
causing, by at least one of the one or more servers of the electronic transaction system, a risk-assessment action to block or allow the current electronic transaction based on the value of impropriety.
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
As per claim 18,
Vashisht explicitly teaches:
further comprising: receiving an indication of a second current electronic transaction associated with an e-commerce platform; based on the indication, providing additional electronic transaction data, corresponding to the second current electronic transaction, to the machine learning model for determining a second value of impropriety for the second current electronic transaction; determining the second value of impropriety is below a threshold; and causing to facilitate electronic payment for the second current electronic transaction based on the second value of impropriety being below the threshold.
(Vashisht US20220261875 at paras. 30-40, 99-102) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0101] In an example as shown in first row, the payment transaction features for a first payment transaction are cross-border, debit card, industry. An authorizing component is applied over the first payment transaction “Product_4” at a time, therefore, a product flag vector of the first payment transaction is 0001. Thus, payment transaction features and product information define a current state of the first payment transaction. The current state will get changed when the issuer 108 applies another authorizing component to the first payment transaction. In another example in the second row, the payment transaction features for a second payment transaction are domestic, credit card, industry. The product flag vector for the second payment transaction is 0010. Further, the industry code vector is a vector representation, where each index value refers to a particular industry type. In the first row, the industry code vector is 01000.")
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. §§ 102 and 103 (or as subject to pre-AIA 35 U.S.C. §§ 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claims 4, 6, 11, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Vashisht, U.S. Patent Application Publication Number 2022/0261875; in view of Faith, U.S. Patent Application Publication Number 2010/0280927; in view of Tseretopoulos, U.S. Patent Application Publication Number 2020/0285644.
As per claim 4,
Vashisht does not explicitly teach, however, Faith does teach:
wherein the plurality of transaction stages for at least one previous electronic transaction comprises [a post-authorization stage] and the stage-differentiated training data comprises delay-captured electronic transaction data identified [after the post-authorization stage].
(Faith US20100280927 at paras. 68-75) ("[0073] In one embodiment, the authorization is only valid during a time window in which the transaction is determined to be likely. In another embodiment, the authorization could specify more than one time window, e.g., if the probability function shows a high likelihood around the 15-17th of the month, but only for 5 pm-10 pm. In yet another embodiment, a pre-authorization sent to a consumer or merchant can continue to be valid for a certain time every week (e.g. 5-7 pm on Thursday). In this manner, a pre-authorization does not have to be sent every week. A revocation of the pre-authorization can be sent when the likelihood of the transaction no longer supports an authorization." "[0140] In one embodiment, calculations for the prediction of an event can be run in real time (e.g. within several hours after an event or series of events occur). In another embodiment, the calculations can be run as batch jobs that are run periodically, e.g., daily, weekly or monthly. For example, a calculation can run monthly to determine who is likely to buy a house, and then a coupon for art, furniture, etc. can be sent to that person. In various embodiments, prediction of major purchases can generally be run in larger batches, whereas prediction of small purchases can be run in real-time (e.g., in reaction to a specific transaction).")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Faith, because it allows for an improved system providing more efficient processing of transactions while maintaining some mechanism for checking fraud. (Faith at Abstract and paras. 3-6).
Vashisht and Faith do not explicitly teach, however, Tseretopoulos does teach:
a post-authorization stage…and…after the post-authorization stage…
(Tseretopoulos US20200285644 at paras. 24-27, 74-76) ("[0024] In some examples, the automated action is at least one of triggered pre-authorization, post-authorization, pre-settlement or post-settlement. [0025] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-authorization, wherein the instruction instructs the transaction processor whether to authorize the associated transaction. [0026] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-settlement, the instruction instructs the transaction processor whether to settle the associated transaction. [0027] In some examples, the records include transaction records each having a status, when the status of a transaction record is post-authorization or post-settlement, the instruction instructs the transaction processor to initialize one or more transactions in relation to the account or one or more related accounts in dependence on the associated transaction." "[0075] At operation 408, when a threshold number of tag selections have been made by the user, the data management device 114 generates a plurality of autotagging rules based on the user selections using, for example, a heuristic learning technique. In some examples, the autotagging rules may be learned by a neural network based on user behavior using, for example, reinforcement learning techniques such as Q-Learning™. Q-Learning™ is a model-free reinforcement learning technique that may be used to find an optimal action-selection policy for any given (finite) Markov decision process (MDP) and is based on learning an action-value function. Using Q-Learning™, an action-value function that automatically determines a tag for a record based on the values of one or more fields of the record may be learned for a particular user and for each account type held by the particular user. The behavior of other users, such as the autotagging rules of other users, may be used as a target in the learning process. Alternatively, the user may define autotagging rules manually by specified the value of one or more fields of the record and the corresponding tag to be assigned. The data management device 114 may suggest autotagging rules based on the autotagging rules of other users from which the user may select one or more autotagging rules.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Faith, and Tseretopoulos, because it provides customization of data categorization which may be more personalized and more precise than alternate approaches such as those which relate on data categorization by data source providers, thereby improving the categorization of data. (Tseretopoulos at Abstract and paras. 2-12).
As per claim 6,
Vashisht explicitly teaches:
wherein the electronic transaction data for each of the previous electronic transactions includes both pre-authorization electronic transaction data and
(Vashisht US20220261875 at paras. 72-75, 105-106) ("[0072] In one embodiment, the data pre-processing engine 220 is configured to randomly select all past transaction-level data (i.e., payment authorization request and payment authorization response messages of past payment transactions) associated with the issuer 108 and/or the merchant 104, for training the RL agent 222. The past transaction-level data associated with the issuer 108 and/or the merchant 104 is stored in the transaction database 118. In other words, the data pre-processing engine 220 is configured to access historical transactions and authorizing components (i.e., products) which were applied by the issuer 108 or the merchant 104 while processing a particular payment transaction. The transaction-level data associated with the issuer 108 or the merchant 104 includes a number of declined/approved/fraud transactions such that the RL agent 222 learns the apt representation of the transaction-level data associated with the issuer 108 or the merchant 104. [0073] In one embodiment, the data pre-processing engine 220 is configured to filter the past transaction-level data with some data constraints (such as, transaction type: card-not-present, decline reason code: addressable declines, issuer and/or merchant geographical region: for example, USA)." "[0105] At 522, the server system 200 accesses historical transaction data associated with the issuer 108 and/or merchant 104. The historical transaction data includes transaction-level data of past payment transaction requests and authorizing component information which were applied to each payment transaction request of the past payment transaction requests. As mentioned above, one or more authorizing components are configured to reduce the decline rates of payment transactions. [0106] At 524, the server system 200 aggregates the past payment transaction requests according to specific data elements. The specific data elements include, but are not limited to, such as, issuer name/identifier, cross border transaction flag, merchant category code (MCC), super industry, month of payment transaction, payment card type, card product type, product flag vector etc. The product flag vector for a particular payment transaction indicates authorizing components (i.e., products) that were enabled on the applied to the particular payment transaction.")
Vashisht and Faith do not explicitly teach, however, Tseretopoulos does teach:
post-authorization electronic transaction data associated with an electronic payment.
(Tseretopoulos US20200285644 at paras. 24-27, 74-76) ("[0024] In some examples, the automated action is at least one of triggered pre-authorization, post-authorization, pre-settlement or post-settlement. [0025] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-authorization, wherein the instruction instructs the transaction processor whether to authorize the associated transaction. [0026] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-settlement, the instruction instructs the transaction processor whether to settle the associated transaction. [0027] In some examples, the records include transaction records each having a status, when the status of a transaction record is post-authorization or post-settlement, the instruction instructs the transaction processor to initialize one or more transactions in relation to the account or one or more related accounts in dependence on the associated transaction." "[0075] At operation 408, when a threshold number of tag selections have been made by the user, the data management device 114 generates a plurality of autotagging rules based on the user selections using, for example, a heuristic learning technique. In some examples, the autotagging rules may be learned by a neural network based on user behavior using, for example, reinforcement learning techniques such as Q-Learning™. Q-Learning™ is a model-free reinforcement learning technique that may be used to find an optimal action-selection policy for any given (finite) Markov decision process (MDP) and is based on learning an action-value function. Using Q-Learning™, an action-value function that automatically determines a tag for a record based on the values of one or more fields of the record may be learned for a particular user and for each account type held by the particular user. The behavior of other users, such as the autotagging rules of other users, may be used as a target in the learning process. Alternatively, the user may define autotagging rules manually by specified the value of one or more fields of the record and the corresponding tag to be assigned. The data management device 114 may suggest autotagging rules based on the autotagging rules of other users from which the user may select one or more autotagging rules.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Faith, and Tseretopoulos, because it provides customization of data categorization which may be more personalized and more precise than alternate approaches such as those which relate on data categorization by data source providers, thereby improving the categorization of data. (Tseretopoulos at Abstract and paras. 2-12).
As per claim 11,
Vashisht explicitly teaches:
wherein the stage-differentiated training data was obtained by applying Markov chain modeling to the previous electronic transaction data for each of the previous electronic transactions to
(Vashisht US20220261875 at paras. 77-79) ("[0078] In order to express the use of reinforcement learning in the product recommendation system for enhancing approval rates of the payment transaction more clearly, the present disclosure explains theoretical models of deep reinforcement learning model, the Markov Decision Process (MDP) with reference to FIG. 4 in more detail. It would be apparent to those skilled in the art that several of deep reinforcement learning models may be applied to accomplish the spirit of the present disclosure." "[0087] FIG. 4 is a block diagram representation of a deep reinforcement learning model 400, in accordance with an embodiment of the present disclosure. As shown in the FIG. 4, the deep reinforcement learning model involves two entities, i.e., an agent 402 (similar to the RL agent 222) and an environment 404, that interacts with each other. The agent 402 is an entity that makes product recommendation decisions, and the environment 404 may be set to feedback a reward value depending upon approval probability and fraud probability scores of a particular transaction and a cost associated with applying a combination of products to the particular transaction. The deep reinforcement learning model 400 implements Markov Decision Process (MDP).")
Vashisht and Faith do not explicitly teach, however, Tseretopoulos does teach:
distinguish pre-authorization electronic payment transaction data from post-authorization electronic transaction data for a post-authorization transaction stage.
(Tseretopoulos US20200285644 at paras. 24-27, 74-76) ("[0024] In some examples, the automated action is at least one of triggered pre-authorization, post-authorization, pre-settlement or post-settlement. [0025] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-authorization, wherein the instruction instructs the transaction processor whether to authorize the associated transaction. [0026] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-settlement, the instruction instructs the transaction processor whether to settle the associated transaction. [0027] In some examples, the records include transaction records each having a status, when the status of a transaction record is post-authorization or post-settlement, the instruction instructs the transaction processor to initialize one or more transactions in relation to the account or one or more related accounts in dependence on the associated transaction." "[0075] At operation 408, when a threshold number of tag selections have been made by the user, the data management device 114 generates a plurality of autotagging rules based on the user selections using, for example, a heuristic learning technique. In some examples, the autotagging rules may be learned by a neural network based on user behavior using, for example, reinforcement learning techniques such as Q-Learning™. Q-Learning™ is a model-free reinforcement learning technique that may be used to find an optimal action-selection policy for any given (finite) Markov decision process (MDP) and is based on learning an action-value function. Using Q-Learning™, an action-value function that automatically determines a tag for a record based on the values of one or more fields of the record may be learned for a particular user and for each account type held by the particular user. The behavior of other users, such as the autotagging rules of other users, may be used as a target in the learning process. Alternatively, the user may define autotagging rules manually by specified the value of one or more fields of the record and the corresponding tag to be assigned. The data management device 114 may suggest autotagging rules based on the autotagging rules of other users from which the user may select one or more autotagging rules.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Faith, and Tseretopoulos, because it provides customization of data categorization which may be more personalized and more precise than alternate approaches such as those which relate on data categorization by data source providers, thereby improving the categorization of data. (Tseretopoulos at Abstract and paras. 2-12).
As per claim 16,
Vashisht does not explicitly teach, however, Faith does teach:
wherein the stage-differentiated training data comprises pre-authorization electronic transaction data distinguished as a first stage of a previous electronic transaction, [post-authorization electronic transaction data distinguished as a second stage of the previous electronic transaction,] and delay-captured electronic transmission data [distinguished as a third stage of the previous electronic transaction for each of the previous electronic transactions by applying Markov chain modeling before training the machine learning model.]
(Faith US20100280927 at paras. 68-75) ("[0073] In one embodiment, the authorization is only valid during a time window in which the transaction is determined to be likely. In another embodiment, the authorization could specify more than one time window, e.g., if the probability function shows a high likelihood around the 15-17th of the month, but only for 5 pm-10 pm. In yet another embodiment, a pre-authorization sent to a consumer or merchant can continue to be valid for a certain time every week (e.g. 5-7 pm on Thursday). In this manner, a pre-authorization does not have to be sent every week. A revocation of the pre-authorization can be sent when the likelihood of the transaction no longer supports an authorization." "[0140] In one embodiment, calculations for the prediction of an event can be run in real time (e.g. within several hours after an event or series of events occur). In another embodiment, the calculations can be run as batch jobs that are run periodically, e.g., daily, weekly or monthly. For example, a calculation can run monthly to determine who is likely to buy a house, and then a coupon for art, furniture, etc. can be sent to that person. In various embodiments, prediction of major purchases can generally be run in larger batches, whereas prediction of small purchases can be run in real-time (e.g., in reaction to a specific transaction).")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Faith, because it allows for an improved system providing more efficient processing of transactions while maintaining some mechanism for checking fraud. (Faith at Abstract and paras. 3-6).
Vashisht and Faith do not explicitly teach, however, Tseretopoulos does teach:
post-authorization electronic transaction data distinguished as a second stage of the previous electronic transaction…and…distinguished as a third stage of the previous electronic transaction for each of the previous electronic transactions by applying Markov chain modeling before training the machine learning model.
(Tseretopoulos US20200285644 at paras. 24-27, 74-76) ("[0024] In some examples, the automated action is at least one of triggered pre-authorization, post-authorization, pre-settlement or post-settlement. [0025] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-authorization, wherein the instruction instructs the transaction processor whether to authorize the associated transaction. [0026] In some examples, the records include transaction records each having a status, when the status of a transaction record is pre-settlement, the instruction instructs the transaction processor whether to settle the associated transaction. [0027] In some examples, the records include transaction records each having a status, when the status of a transaction record is post-authorization or post-settlement, the instruction instructs the transaction processor to initialize one or more transactions in relation to the account or one or more related accounts in dependence on the associated transaction." "[0075] At operation 408, when a threshold number of tag selections have been made by the user, the data management device 114 generates a plurality of autotagging rules based on the user selections using, for example, a heuristic learning technique. In some examples, the autotagging rules may be learned by a neural network based on user behavior using, for example, reinforcement learning techniques such as Q-Learning™. Q-Learning™ is a model-free reinforcement learning technique that may be used to find an optimal action-selection policy for any given (finite) Markov decision process (MDP) and is based on learning an action-value function. Using Q-Learning™, an action-value function that automatically determines a tag for a record based on the values of one or more fields of the record may be learned for a particular user and for each account type held by the particular user. The behavior of other users, such as the autotagging rules of other users, may be used as a target in the learning process. Alternatively, the user may define autotagging rules manually by specified the value of one or more fields of the record and the corresponding tag to be assigned. The data management device 114 may suggest autotagging rules based on the autotagging rules of other users from which the user may select one or more autotagging rules.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Faith, and Tseretopoulos, because it provides customization of data categorization which may be more personalized and more precise than alternate approaches such as those which relate on data categorization by data source providers, thereby improving the categorization of data. (Tseretopoulos at Abstract and paras. 2-12).
Claims 8, 12, 19, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Vashisht, U.S. Patent Application Publication Number 2022/0261875; in view of Chen, U.S. Patent Application Publication Number 2024/0095742.
As per claim 8,
Vashisht does not explicitly teach, however, Chen does teach:
wherein the first electronic transaction is blocked during the pre-authorization transaction stage for the first electronic transaction, and wherein the computer-implemented method further comprises: updating the machine learning model by rewarding the machine learning model based on blocking the first electronic transaction during the pre-authorization transaction stage.
(Chen US20240095742 at paras. 29-33, 47-53) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
As per claim 12,
Vashisht does not explicitly teach, however, Chen does teach:
wherein the neural network is updated using a reward function and reinforcement learning, such that the neural network is rewarded for blocking the current electronic transaction based on the value of impropriety.
(Chen US20240095742 at paras. 29-33, 47-53) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
As per claim 19,
Vashisht does not explicitly teach, however, Chen does teach:
further comprising causing application of reinforcement learning to the machine learning model in response to determining the value of impropriety and the second value of impropriety based on a reward for blocking the current electronic transaction and a second reward for facilitating the electronic payment for the second current electronic transaction, the reward for blocking being a greater reward than the second reward.
(Chen US20240095742 at paras. 29-33, 47-63) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments." "[0053] In a further example embodiment, a sampling strategy is used to sample episodes rather than using all of them. Specifically, episodes may be downsampled where (1) all events are true negatives; or (2) all events are observed as false positives (blocked by the decline model 310 but nevertheless allowed to proceed) and none were subject to a chargeback. [0054] As mentioned above, in an example embodiment the decision model 314 attempts to optimize on revenues. In other words, given a transaction described by a set of features, the decision model 314 aims to maximize cumulative rewards by choosing an appropriate action. The action can be to allow or block the transaction, but can also include future interventions as well (e.g., blocking future transactions from this customer). The reward describes the expected profit or loss as a result of the chosen action, and can also include just the reward from the current transaction or future rewards as well.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
As per claim 20,
Vashisht does not explicitly teach, however, Chen does teach:
further comprising causing application of reinforcement learning to the machine learning model based on a punishment upon the machine learning model providing the value of impropriety that is below a threshold for a fraudulent electronic transaction.
(Chen US20240095742 at paras. 29-33, 47-63) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments." "[0053] In a further example embodiment, a sampling strategy is used to sample episodes rather than using all of them. Specifically, episodes may be downsampled where (1) all events are true negatives; or (2) all events are observed as false positives (blocked by the decline model 310 but nevertheless allowed to proceed) and none were subject to a chargeback. [0054] As mentioned above, in an example embodiment the decision model 314 attempts to optimize on revenues. In other words, given a transaction described by a set of features, the decision model 314 aims to maximize cumulative rewards by choosing an appropriate action. The action can be to allow or block the transaction, but can also include future interventions as well (e.g., blocking future transactions from this customer). The reward describes the expected profit or loss as a result of the chosen action, and can also include just the reward from the current transaction or future rewards as well.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
Claims 13 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Vashisht, U.S. Patent Application Publication Number 2022/0261875; in view of Chen, U.S. Patent Application Publication Number 2024/0095742; in view of Aparício, U.S. Patent Application Publication Number 2021/0142329.
As per claim 13,
Vashisht explicitly teaches:
during a pre-authorization transaction stage for the current electronic transaction based on a pre-authorization value of impropriety determined by the neural network, and wherein the neural network is updated using reinforcement learning,
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
Vashisht does not explicitly teach, however, Chen does teach:
the reinforcement learning includes a punishment upon the neural network for not blocking the current electronic transaction during the pre-authorization transaction stage based on the pre-authorization value of impropriety.
(Chen US20240095742 at paras. 29-33, 47-53) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments." "[0053] In a further example embodiment, a sampling strategy is used to sample episodes rather than using all of them. Specifically, episodes may be downsampled where (1) all events are true negatives; or (2) all events are observed as false positives (blocked by the decline model 310 but nevertheless allowed to proceed) and none were subject to a chargeback.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
Vashisht and Chen do not explicitly teach, however, Aparício does teach:
wherein the current electronic transaction is fraudulent and was not blocked
(Aparício US20210142329 at paras. 50-60) ("[0055] The process begins by receiving a specification of past predicted results of evaluation rules and corresponding observed outcomes (400). The specification of past predicted results of evaluation rules includes a set of evaluation rules and predicted results of the evaluation rules. The set of evaluation rules includes one or more rules that can be applied to data to make predictions about the data. For example, in a fraud detection system, the evaluation rules are applied to transactional data to predict whether a transaction is fraudulent or not. The past predicted results of the evaluation rules refer to the actions/determinations made by the evaluation rules on data from the past and for which there are observed actual outcomes indicating whether the transaction turned out to be fraudulent or not. The types of actions/determinations made by the evaluation rules may be domain-specific. Referring again to the fraud detection case, actions may include accept a transaction, decline a transaction, or generate an alert to perform further review of a transaction. The observed outcomes may be obtained in a variety of ways including by conventional methods such as assessment by a security analyst or data scientist who determines a true label for a transaction. [0056] In various embodiments, the received specification includes an assessment of the performance of the set of evaluation rules. Alternatively or in addition, the process assess the received specification. The performance of evaluation rules (a rule set or portion thereof) may be assessed by comparing the past predicted results with observed outcomes. A rule set that performs well would predict results that are similar to the observed outcomes, while a rule set that performs poorly would predict results that are different from the observed outcomes (false positives and false negatives). For example, a rule set that does not perform well would accept a transaction when the observed outcome is a chargeback. In other words, rule sets that tend to generate more (above some threshold number) false positives or false negatives perform relatively poorly. For example, in a fraud case, the threshold may be based on a cost function that maximizes recall subject to not increasing the number of false positives (e.g., the FPR threshold is below 10%, the FPR threshold increases less than some amount such as no more than 10% where it used to be 5%, or the FPR remains the same as before). As further described below, rule set performance can be assessed in various ways and assigned a score or category (e.g., good, fair, poor). An example evaluation process is shown in FIG. 8.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Chen, and Aparício, because it improves upon conventional complex rule systems by making them more effective and ensures that performance objectives are met in a fraud detection system. (Aparício at Abstract and paras. 2, 42).
As per claim 14,
Vashisht explicitly teaches:
during a pre-authorization transaction stage for the current electronic transaction based on a pre-authorization value of impropriety determined by the neural network, and wherein the neural network is updated using reinforcement learning,
(Vashisht US20220261875 at paras. 30-40, 99-102, 130-140) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0134] At 616, the server system 200 checks whether the reward value is greater than or equal to a threshold value or not. [0135] At 618, when the reward value is greater than or equal to the threshold value, the server system 200 adds the candidate authorizing component into a product recommendation strategy. [0136] At 620, when the reward value is not greater than the threshold value, the server system 200 selects another action (e.g., application of another candidate authorizing component to the payment transaction) that has a maximum Q-value from all Q-values for all possible actions in the new state. [0137] At 622, the server system 200 transmits the payment authorization request along with the product recommendation strategy to the issuer in the real-time. The issuer 108 applies one or more authorizing components included in the product recommendation strategy to the payment transaction, resulting in high approval rates, lower fraud risk and maximized revenues for issuers and/or merchants.")
Vashisht does not explicitly teach, however, Chen does teach:
the reinforcement learning includes a punishment, upon the neural network for blocking the current electronic transaction during the pre-authorization transaction stage based on the pre-authorization value of impropriety.
(Chen US20240095742 at paras. 29-33, 47-53) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments." "[0053] In a further example embodiment, a sampling strategy is used to sample episodes rather than using all of them. Specifically, episodes may be downsampled where (1) all events are true negatives; or (2) all events are observed as false positives (blocked by the decline model 310 but nevertheless allowed to proceed) and none were subject to a chargeback.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
Vashisht and Chen do not explicitly teach, however, Aparício does teach:
wherein the current electronic transaction is non-fraudulent and was blocked
(Aparício US20210142329 at paras. 50-60) ("[0055] The process begins by receiving a specification of past predicted results of evaluation rules and corresponding observed outcomes (400). The specification of past predicted results of evaluation rules includes a set of evaluation rules and predicted results of the evaluation rules. The set of evaluation rules includes one or more rules that can be applied to data to make predictions about the data. For example, in a fraud detection system, the evaluation rules are applied to transactional data to predict whether a transaction is fraudulent or not. The past predicted results of the evaluation rules refer to the actions/determinations made by the evaluation rules on data from the past and for which there are observed actual outcomes indicating whether the transaction turned out to be fraudulent or not. The types of actions/determinations made by the evaluation rules may be domain-specific. Referring again to the fraud detection case, actions may include accept a transaction, decline a transaction, or generate an alert to perform further review of a transaction. The observed outcomes may be obtained in a variety of ways including by conventional methods such as assessment by a security analyst or data scientist who determines a true label for a transaction. [0056] In various embodiments, the received specification includes an assessment of the performance of the set of evaluation rules. Alternatively or in addition, the process assess the received specification. The performance of evaluation rules (a rule set or portion thereof) may be assessed by comparing the past predicted results with observed outcomes. A rule set that performs well would predict results that are similar to the observed outcomes, while a rule set that performs poorly would predict results that are different from the observed outcomes (false positives and false negatives). For example, a rule set that does not perform well would accept a transaction when the observed outcome is a chargeback. In other words, rule sets that tend to generate more (above some threshold number) false positives or false negatives perform relatively poorly. For example, in a fraud case, the threshold may be based on a cost function that maximizes recall subject to not increasing the number of false positives (e.g., the FPR threshold is below 10%, the FPR threshold increases less than some amount such as no more than 10% where it used to be 5%, or the FPR remains the same as before). As further described below, rule set performance can be assessed in various ways and assigned a score or category (e.g., good, fair, poor). An example evaluation process is shown in FIG. 8.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Chen, and Aparício, because it improves upon conventional complex rule systems by making them more effective and ensures that performance objectives are met in a fraud detection system. (Aparício at Abstract and paras. 2, 42).
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Vashisht, U.S. Patent Application Publication Number 2022/0261875; in view of Faith, U.S. Patent Application Publication Number 2010/0280927; in view of Tseretopoulos, U.S. Patent Application Publication Number 2020/0285644; in view of Chen, U.S. Patent Application Publication Number 2024/0095742.
As per claim 17,
Vashisht explicitly teaches:
wherein the current electronic payment transaction is blocked during the first stage.
(Vashisht US20220261875 at paras. 30-40, 99-102) ("[0033] Various example embodiments of the present disclosure provide methods, systems, user devices and computer program products for enhancing approval rates of payment processing requests by recommending application of one or authorizing components to payment transactions to issuers, in real time. The one or more authorizing components help the issuers in taking authorization decisions and are configured to decline fraud transactions. The system determines which all authorizing components issuers and/or merchants should apply on a particular transaction so as to maximize the approval probability while minimizing the fraud probability. The system also optimizes the cost of applying an authorizing component to the payment transaction for the issuers and/or merchants." "[0038] Thereafter, the server system is configured to calculate Q-values corresponding to state-action pairs formed by the state and the actions using a neural network of the deep reinforcement learning model. The server system is configured to select an action (i.e., application of a candidate authorizing component to the payment transaction) based at least on the calculated Q-values and epsilon greedy policy methods. The server system is configured to calculate a reward value corresponding to the selected action based, at least in part, on a reward function. The reward function is based on approval and fraud probability scores of a payment transaction type associated with the payment transaction and a cost of applying the selected candidate authorizing component to the payment transaction." "[0101] In an example as shown in first row, the payment transaction features for a first payment transaction are cross-border, debit card, industry. An authorizing component is applied over the first payment transaction “Product_4” at a time, therefore, a product flag vector of the first payment transaction is 0001. Thus, payment transaction features and product information define a current state of the first payment transaction. The current state will get changed when the issuer 108 applies another authorizing component to the first payment transaction. In another example in the second row, the payment transaction features for a second payment transaction are domestic, credit card, industry. The product flag vector for the second payment transaction is 0010. Further, the industry code vector is a vector representation, where each index value refers to a particular industry type. In the first row, the industry code vector is 01000.")
Vashisht, Faith, and Tseretopoulos does not explicitly teach, however, Chen does teach:
wherein the previous electronic transactions used for training the machine learning model include both fraudulent and non-fraudulent previous electronic transactions, and
(Chen US20240095742 at paras. 29-33, 47-63) ("[0048] High fraud tolerance segments may include, for example, merchants that are funded startups, in early stages, or in high margin industries. Low fraud tolerance segments may include, for example, merchants in low margin industries such as groceries, food and drink, and deliveries. [0049] The decision model 314 may be trained to classify a merchant into a fraud tolerance segment based partially on the merchant's explicit preference but also partially based on maximizing a revenue optimization function. This may be accomplished using the historical transaction data 315, which can be utilized during the training of the decision model 314 to select parameters for merchants based on merchant information, the parameters being selected based on which parameters would maximize revenue for the corresponding merchant. This allows fraud tolerance parameters used by a particular merchant to influence fraud tolerance parameters used by another, similar, merchant. For example, merchant A may have a significant amount of historical transaction data 315, which may be used during the training of the decision model 314 to assign values to parameters that cause the fraud tolerance threshold for merchant A to be set to a level that maximizes revenue overall for merchant A (at least for the historical transaction data 315 it was trained on). At runtime, however, a transaction may be evaluated for merchant B, who may not have had historical transaction data 315 used to train the decision model 314, or at least had less historical transaction data 315 such that the prediction by the decision model might have otherwise been unreliable if it were limited to using only merchant B's historical transaction data 315. Nevertheless, since merchant B is similar (e.g., similar industry, funding status, etc.) to merchant A, the decision model 314 may utilize the parameters from merchant A for merchant B, or at least have such parameters influence the parameters for merchant B. Put another way, the decision model 314 classifies merchant B as being in the same fraud tolerance segment as merchant A based on information known about merchant B and the merchants in the various fraud tolerance segments." "[0053] In a further example embodiment, a sampling strategy is used to sample episodes rather than using all of them. Specifically, episodes may be downsampled where (1) all events are true negatives; or (2) all events are observed as false positives (blocked by the decline model 310 but nevertheless allowed to proceed) and none were subject to a chargeback. [0054] As mentioned above, in an example embodiment the decision model 314 attempts to optimize on revenues. In other words, given a transaction described by a set of features, the decision model 314 aims to maximize cumulative rewards by choosing an appropriate action. The action can be to allow or block the transaction, but can also include future interventions as well (e.g., blocking future transactions from this customer). The reward describes the expected profit or loss as a result of the chosen action, and can also include just the reward from the current transaction or future rewards as well.")
Therefore, it would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Vashisht, Faith, Tseretopoulos, and Chen, because it improves the use of machine learning for fraud tolerance in online networks. (Chen at Abstract and paras. 1-15).
Response to Arguments
Applicant’s arguments filed on March 2, 2026 have been fully considered but are not persuasive for the following reasons:
With respect to Applicant’s arguments as to the § 101 rejections for now pending claims 1-20, Examiner notes the following:
Applicant argues that the amended features would integrate the abstract idea into a practical application, the examiner respectfully disagrees. In particular, the applicant argues that the claims integrate “any such purported idea into a practical application by reciting a particular, technically-rooted way of improving how electronic transaction systems detect and act on fraud risk across multiple transaction checkpoints using stage-aware model training and stage-specific reward updating” and “The improvement is to the technology of multi-stage fraud/risk detection in electronic transaction systems, because the claim's stage-aware training and stage-aware inference are directed at enhancing accuracy and operational effectiveness across checkpoints while avoiding the drawbacks of conventional designs.”
Examiner disagrees, however, and notes that, the additional elements of the computer system - a “A computer-implemented method comprising:”, “at least one of one or more servers of an electronic transaction system”, “A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising:”, “neural network”, and “One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising:” to perform the steps of “accessing”, “processing”, “associating”, “training”, “employing”, “determining”, “blocking”, and “providing”, in all steps is recited at a high-level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer component. The claims at issue collecting, analyzing, and transmitting data to determine whether to block or allow a transaction, e.g., associated with a payment transaction. The claims invoke the “A computer-implemented method comprising:”, “at least one of one or more servers of an electronic transaction system”, “A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising:”, “neural network”, and “One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising:” to perform the steps of “accessing”, “processing”, “associating”, “training”, “employing”, “determining”, “blocking”, and “providing” merely as tools to execute the abstract idea. Use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a certain method of organizing human activity or mental process or mathematical calculation) does not integrate a judicial exception into a practical application. (MPEP 2106.05 (f))
Finally, the Applicant argues that the claims are directed to significantly more than the abstract idea.
Examiner disagrees, however, and notes that, as explained above in the instant rejection under 35 U.S.C. § 101, that the additional elements do not amount to an inventive concept. The additional elements of the computer system - “A computer-implemented method comprising:”, “at least one of one or more servers of an electronic transaction system”, “A computer system comprising: one or more processors; and a computer storage medium storing computer-useable instructions that, when used by the one or more processors, causes the computer system to perform operations comprising:”, “neural network”, and “One or more non-transitory computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations comprising:” to perform the steps of “accessing”, “processing”, “associating”, “training”, “employing”, “determining”, “blocking”, and “providing” are merely generic computer components performing their well-known basic functions of collecting, analyzing, and transmitting data to determine whether to block or allow a transaction, e.g., associated with a payment transaction. Per the specification, the recited computer elements and machine learning steps and model are described only at a high level of generality, (see Spec. at paras. [0024], [0034]). In view of the specification, the application of the computer elements and machine learning model is merely being applied to the abstract idea.
The other limitations which are simply supporting the abstract idea correspond to insignificant extra-solution activity which do not transform the abstract idea into a patent eligible subject matter. Also, the functionality here is already present in the recited hardware, which is merely routine and conventional. Collecting, analyzing, and transmitting data is routine and conventional. There is no technological problem or solution identified. This is merely a business solution to transfer data between devices. (MPEP 2106.05 (f))
With respect to Applicant’s arguments as to the § 103 rejections for now pending claims 1-20, Examiner notes that the arguments are moot in light of the new grounds for rejection.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure and is available for review on Form PTO-892 Notice of References Cited.
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 extension fee 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 date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MERRITT J HASBROUCK whose telephone number is (571)272-3109. The examiner can normally be reached M-F 9:00-5:00.
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, Christine Tran can be reached on 571-272-8103. 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.
/MERRITT J HASBROUCK/Examiner, Art Unit 3695
/CHRISTINE M Tran/Supervisory Patent Examiner, Art Unit 3695
Prosecution Insights