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 .
Examiner Notes
Examiner cites particular columns and line numbers in the references as applied to the claims below for convenience of the applicant. Although the specified citations are representative of the teachings in the art and are applied to the specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested that, in preparing responses, the applicant fully consider the references cited in their entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner.
Response to Amendment
The Amendment filed 7/20/2026 has been entered. Claims 1-20 remain pending in the present Office Action.
The Amendments to the claims have been fully considered and overcome all Objections and Rejections under 35 U.S.C. 112(b) set forth in the previous Office Action.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites: A method, comprising:
(a) receiving an event identifier and historical data for at least one event;
(b) calculating an event frequency of the at least one event;
(c) identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein a different models are associated with different event frequency designations;
(d) training the first model based on the historical data for the at least one event, wherein training the first model based on the historical data for the at least one event further comprises:
(e) identifying at least one event change point in the historical data; and
(f) calculating an event time slot based on the at least one event change point in the historical data; and
(g) generating a prediction of one or more predicted future events based at least in part on the first model.
Step 1:
Claim 1 is reciting a method, which is a process claimed as a series of steps, which is one of the four statutory categories.
Step 2A, Prong One:
Steps (b), (c), (d), (e), (f), and (g) are all reciting steps which can be performed in the human mind, with the aid of pen and paper, through observation, evaluation, judgement and opinion.
Specifically, a person is capable of performing step (b) calculating a frequency an event (e.g. a deposit in a bank account) through observation of how often the event occurs (e.g., through looking at their account history).
A person is capable of performing step (c) identifying a model based on the calculated frequency, where different models are associated with different frequency designations, through judgment (e.g., a person is capable of determining the calculated frequency falls into a rare frequency designation, and identify a model, e.g. a rules-based model, associated with a rare frequency designation should be used).
A person is capable of performing step (d) training a model through evaluation. Specifically, “training a model” as is known in the art encompasses an iterative process of feeding data to a model such that it generates a result, which can be compared to an actual known result, and adjusting the model based on the differences between the generated result and the known result to reduce the difference. Since a “model” as claimed is not limited to any particular type of model, a model could comprise any logic that simulates/represents a real-world system or process, e.g., to assist in making a prediction. A person would be capable of performing logic on (applying a set of rules to) historical event data to classify a day as having an event (e.g., a paycheck deposit) or not having an event, comparing the results of the logic to the actual data (e.g., comparing the days which the rules predicted a deposit vs. when the deposits were actually received), and adjusting the logic (e.g., adding or changing a rule in the logic) to improve the accuracy of the model.
A person is capable of performing step (e) identifying an event change point in historical data through observation (e.g., identifying when the day for receiving a paycheck deposit changes from Friday to Thursday).
A person is capable of performing step (f) calculating an event time slot based on the event change point through observation (e.g., determining Thursday is when a paycheck deposit should be received after it was identified that the day for receiving the paycheck changed from Friday to Thursday).
A person is capable of performing step (g) generating a prediction of one or more predicted future events based at least in part on the first model through judgement and evaluation (e.g., a person could predict they will receive a paycheck deposit on a particular Thursday based on an output of a model).
Accordingly, claim 1 recites a judicial exception (i.e., an abstract idea).
Step 2A, Prong Two:
The only additional element recited by claim 1 is step (a).
Regarding step (a), the limitation recited is insignificant extra-solution activity of necessary data gathering. This additional element of insignificant extra-solution activity is not indicative of integration into a practical application. See MPEP 2106.04(d) and 2106.05(g).
Even when considering the claim as a whole, the claim amounts to gathering data and a plurality of steps performed on the gathered data which can be performed in the human mind. Thus, the judicial exceptions are not integrated into a practical application. See MPEP 2106.04(d).
Step 2B:
Regarding the additional element recited in step (a), the limitation amounts to insignificant extra-solution activity of mere data gathering. Further, the additional element recited in step (a) is well-understood, routine and conventional activity of receiving data, e.g., over a network, as identified in MPEP 2106.05(d). The courts have found that adding insignificant extra-solution activity and well-understood, routine, conventional activity is not enough to amount to significantly more than the recited judicial exception. See MPEP 2106.05.
Even when considering the claim as a whole, the claim amounts to a step of insignificant extra-solution activity, which is also a well-understood, routing and conventional activity, of gathering data and a plurality of steps performed on the data which can be performed in the human mind. Consequently, claim 1 as a whole does not amount to significantly more than the judicial exception and the claim is not eligible.
Claim 2 is dependent on claim 1, and therefore inherits the same judicial exceptions recited in claim 1. Further, claim 2 recites the limitations: monitoring for the one or more predicted future events based on the calculated event frequency of the at least one event; and detecting when a given predicted future event of the one or more predicted future events does not occur. These limitations can be performed mentally, with the aid of pen and paper, through observation, evaluation, judgement and opinion, and are therefore reciting a judicial exception of an abstract idea (i.e., a mental process).
Claim 2 does not recite any additional elements beyond those recited in claim 1. Accordingly, the additional elements are not indicative of integration into a practical application, nor do they provide an inventive concept. Thus, claim 2 is not eligible.
Claim 3 is dependent on claim 2, and therefore inherits the same judicial exceptions recited in claims 1 and 2. Further, claim 3 recites the limitation: initiating an alert when the given predicted future event of the one or more predicted future events does not occur. This limitation can be performed mentally, with the aid of pen and paper (e.g., a person deciding to send a message or speak to their employer when a paycheck is not deposited as predicted is “initiating an alert”), through observation, evaluation, judgement and opinion, and is therefore reciting a judicial exception of an abstract idea (i.e., a mental process).
Claim 3 does not recite any additional elements beyond those recited in claim 1. Accordingly, the additional elements are not indicative of integration into a practical application, nor do they provide an inventive concept. Thus, claim 3 is not eligible.
Claim 4 is dependent on claim 1, and therefore inherits the same judicial exceptions recited in claim 1. Further, claim 4 recites the limitations wherein an event frequency comprises: one of daily, weekly, monthly quarterly, or yearly. These limitations (“calculating an event frequency of the at least one event […] wherein an event frequency comprises: one of daily, weekly, monthly quarterly, or yearly”) can be performed mentally, with the aid of pen and paper, through observation, evaluation, judgement and opinion, and are therefore reciting a judicial exception of an abstract idea (i.e., a mental process).
Claim 4 does not recite any additional elements beyond those recited in claim 1. Accordingly, the additional elements are not indicative of integration into a practical application, nor do they provide an inventive concept. Thus, claim 4 is not eligible.
Claim 5 is dependent on claim 4, and therefore inherits the same judicial exceptions recited in claims 1 and 4. Further, claim 5 recites the limitation wherein calculating the event frequency comprises:
calculating a mean frequency for the event. This limitation can be performed mentally, with the aid of pen and paper, through observation, evaluation, judgement and opinion, and is therefore reciting a judicial exception of an abstract idea (i.e., a mental process). Further, this limitation is also reciting an abstract idea of a mathematical concept (calculating a mean is a mathematical calculation).
Claim 5 does not recite any additional elements beyond those recited in claim 1. Accordingly, the additional elements are not indicative of integration into a practical application, nor do they provide an inventive concept. Thus, claim 5 is not eligible.
Claim 6 is dependent on claim 1, and therefore inherits the same judicial exceptions recited in claim 1.
Claim 6 recites the additional element of wherein the first model of the plurality of models comprises one of: a seasonality model, a sequence model, a transformer model, a statistical model, or a rules-based model, which amounts to generally linking a particular technological environment (e.g., to a machine learning environment, in which a machine learning model may be trained and used in generating a prediction).
This additional element of generally linking to a particular technological environment is not indicative of integration into a practical application. Even when considered in combination with the additional elements of claim 1, the additional elements recited comprise insignificant extra-solution activity and generally linking the use of an exception to a particular technological environment, which fail to integrate the recited judicial exceptions into a practical application. See MPEP 2106.04(d). Further, these additional elements of insignificant extra-solution activity and generally linking the use of an exception to a particular technological environment do not amount to significantly more than the recited exceptions. See MPEP 2106.05. Thus, claim 6 is not eligible.
Claim 7 is dependent on claim 1, and therefore inherits the same judicial exceptions recited in claim 1. Further, claim 7 recites the limitation: wherein generating a prediction of one or more predicted future events based at least in part on the first model comprises: estimating at least one event time within the calculated event time slot for the one or more predicted future events. This limitation can be performed mentally, with the aid of pen and paper, through observation, evaluation, judgement and opinion, and is therefore reciting a judicial exception of an abstract idea (i.e., a mental process).
Claim 7 does not recite any additional elements beyond those recited in claim 1. Accordingly, the additional elements are not indicative of integration into a practical application, nor do they provide an inventive concept. Thus, claim 7 is not eligible.
Claim 8 is dependent on claim 1, and therefore inherits the same judicial exceptions recited in claim 1. Further, claim 8 recites the limitations wherein the event frequency designations comprise one of: a frequent event, a moderate event, or a rare event. These limitations (“identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein a different models are associated with different event frequency designations; […] wherein the event frequency designations comprise one of: a frequent event, a moderate event, or a rare event”) can be performed mentally, with the aid of pen and paper, through observation, evaluation, judgement and opinion, and are therefore reciting a judicial exception of an abstract idea (i.e., a mental process).
Claim 8 does not recite any additional elements beyond those recited in claim 1. Accordingly, the additional elements are not indicative of integration into a practical application, nor do they provide an inventive concept. Thus, claim 8 is not eligible.
Claims 9-16 are directed to A system, comprising: a computer having a processor and a memory; and one or more code sets stored in the memory and executed by the processor, which, when executed, configure the processor to: perform the methods of claims 1-8, respectively. As such, in view of the abovementioned reasons presented with respect to claims 1-8, claims 9-16 are also directed to a judicial exception without significantly more and are ineligible.
For clarity, the additional element of claim 9 recited above is considered to be mere instructions to apply the exception, which is not indicative of integration into a practical application nor does it amount to significantly more than the judicial exception.
Claims 17-20 are directed to A non-transitory computer-readable medium storing computer-program instructions that, when executed by one or more processors, cause the one or more processors to effectuate operations comprising: perform the methods of claims 1-3 and 8, respectively. As such, in view of the abovementioned reasons presented with respect to claims 1-3 and 9, claims 17-20 are also directed to a judicial exception without significantly more and are ineligible.
For clarity, the additional element of claim 17 recited above is considered to be mere instructions to apply the exception, which is not indicative of integration into a practical application nor does it amount to significantly more than the judicial exception.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 4-9, 12-17 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Neznal (U.S. Pub. No. 2020/0380335), in view of Shashikant Rao et al. (U.S. Pub. No. 2020/0242483), hereinafter Shashikant Rao, and Garvey et al. (U.S. Pub. No. 2017/0249564), hereinafter Garvey.
Regarding claim 1, Neznal teaches a method, comprising:
receiving an event identifier and historical data for at least one event ([0027] – “a series of commercial transactions is broken down into a time series of high-dimensional inputs, where the dimensions are features of the commercial transactions such as the country or zip code of a transaction or an item code identifying the product or service provided. These high-dimensional inputs are input to the LSTM neural network in windowed sequences, both to train the network and subsequently to evaluate commercial transactions for anomalies.”; [0028] – “training data historic commercial transaction data 116 […] The commercial transaction data 116 includes data from a variety of normal commercial transactions that can be used to train the recurrent neural network, and in a further example includes anomalous commercial transactions that can be used to help train the neural network to better identify anomalies.”; [0030] – “receiving new commercial transaction data as it is recorded in 116, or by periodically processing new commercial transaction data from commercial transaction database 116.”; [0033] – “Time series are created for various combinations of measures such as bookings, installations, sales, etc. , and measures such as physical location, software version or other product identifier, number of products or services bought, identity of each product or service bought, etc.”; [0034] – “The preprocessed high-dimensional data is provided to the recurrent neural network in a time series” Historical data of a plurality of transactions (i.e., “events”) is received. This historical data for each transaction may comprise an item code/some other identifier of the product/service bought in the transaction (i.e., an “event identifier”).);
calculating an event frequency of the at least one event ([0027] – “commercial transaction features are compiled per hour, per day, or over other time periods during which commercial transactions are observed to be similar or have repeating patterns.” Compiling transactions, e.g., X transactions, per time period, e.g., per hour, is calculating the frequency of the transactions (“events”), e.g., the transaction frequency = X transactions/hour.); […]
training the first model based on the historical data for the at least one event ([0028] – “training data historic commercial transaction data 116 […] The commercial transaction data 116 includes data from a variety of normal commercial transactions that can be used to train the recurrent neural network”), wherein training the first model based on the historical data for the at least one event further comprises: […]
calculating an event time slot based on […] the historical data ([0041]-[0042] – “the window size for the one-dimensional input shown is five records, such as five hours of commercial transactions, but in many other examples will be longer, such as a day, week, or months' worth of commercial transactions. These overlapping sequences are extracted are therefore each the same size, extracted from the time series of observed commercial transaction data. These windowed time series of data are provided to the network during training with the knowledge of the next element in the data series outside the input window, which is used to train the recurrent neural network to predict the next data element. In operation, the windowed data is provided as an input to the recurrent neural network to generate a predicted output”; [0052] – “At 1106, the high-dimensional time series is windowed, such as by taking sequential overlapping groups of the time series, incremented by a time over which patterns are likely to repeat such as a day, a week, or a month, and provided to the recurrent neural network for training. The time series window is evaluated at 1108 to generate or output a predicted next element or elements in the series”; [0032] – “high-dimensional space further includes time-based information such as day of the week or hour of the day during which a transaction is conducted”); and
generating a prediction of one or more predicted future events based at least in part on the first model ([0034] – “The preprocessed high-dimensional data is provided to the recurrent neural network in a time series […] The recurrent neural network process the data in a manner that uses both prior state data and current state data to predict the next data likely to be observed in the commercial transaction data series”; [0035] – “FIG. 2 is a chart showing use of a trained recurrent neural network to identify commercial transaction anomalies, consistent with an example embodiment. Here, the predicted occurrence of an event, such as revenue of a product or other such event characterized by the high-dimensional preprocessing of the network data, is charted.”; [0054] – “the high-dimensional time series windowed input is evaluated to generate an output of a predicted next element or elements in the series.”).
Neznal fails to expressly teach identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations and training the model comprises: identifying at least one event change point in the historical data; and calculating the event time slot based on the at least one event change point in the historical data.
However, Shashikant Rao teaches identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations ([0028]-[0029] – “each time series data set comprises varying levels of sparsity. As a specific illustrative example, a time series data set of sales may be relatively dense due to a high volume of sales transactions, such as sales at a grocery store. As a specific illustrative example, a time series data set of payroll transactions may be moderately sparse due to a moderate volume of paychecks that are provided to employees every two weeks. As a specific illustrative example, a time series data set of government regulation transactions may be severely sparse due to a low volume of payments to regulatory entities that occur once a quarter. It is to be understood that the infrequency of transactions of a time series can impact the sparsity of the time series. In one embodiment, the meta-model training system 120 models financial transactions as historical time series data sets. In one embodiment, the service provider computing environment 110 utilizes the historical time series data sets and forecasts those data sets into the future. Such forecasts are beneficial to users of the service provider computing environment 110 to provide insights into events that may occur in the future”; [0032] – “the time series data sets have varying levels of sparsities. As a specific illustrative example, a time series is dense when transactions are generated hourly and is sparse when transactions are generated weekly. In one embodiment, time series are grouped by sparsity levels as discussed further below with respect to FIGS. 7 and 8. In one embodiment, an appropriate model is chosen based on a sparsity level grouping. In one embodiment, sparsity level groupings define subsets of time series. As a specific illustrative example, a first grouping may be low sparsity, a second grouping may be medium sparsity, and a third grouping may be high sparsity.”; [0040] – “time series data comprises an event stream. In one embodiment, an event is represented as a data point occurring in time, and an event stream is a stream of events”; [0131]-[0132] – a time series with a high sparsity property has fewer data points over time, due to the type of transaction for the time series being less frequent; likewise, transactions that are more frequent produce a time series with lower sparsity).
Neznal and Shashikant Rao are considered to be pertinent prior art to the claimed invention because they are in the same field of predicting future events using a model. Therefore, it would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal to incorporate the teachings of Shashikant Rao such that the model used to generate a prediction of future events is selected from a plurality of models based on the calculated frequency of the at least one event. Using different models for time series data with different sparsities (e.g., due to different event frequencies) allows for faster and more accurate predictions than what could be predicted by a single algorithm (Shashikant Rao: [0002], [0005], and [0039]).
The combination of Neznal in view of Shashikant Rao fails to expressly teach training the model comprises: identifying at least one event change point in the historical data; and calculating the event time slot based on the at least one event change point in the historical data.
However, Garvey teaches training a prediction model comprises: identifying at least one event change point in the historical data ([0034] – “Systems and methods are described through which state changes within a time-series signal may be automatically detected and accommodated in a forecast or other analytical model. A "state change" in this context refers to a change in the "normal" characteristics of the signal. For instance, time-series data before a particular point in time (a "change time") may exhibit a first pattern or set of one or more characteristics that are determined to be normal. After the change time, the time-series data may exhibit a second pattern or second set of characteristics that were not normal before the change time. In other words, a state change results in a "new normal" for the time-series signal.”; [0080] – “Once detected, the state changes are further analyzed to determine whether they are normal or abnormal. Changes that are normal do not cause a "new" normal in the state of the time-series signal”; [0088] – “At 710, the process identifies the most recent state change that is not normal.”; [0053] – time series signal may measure events, e.g., database logons); and calculating an event time slot based on the at least one event change point in the historical data ([0086] – “time-series data that occurs before a change time corresponding to an abnormal change may be excluded from the training set while time-series data occurring after the change time may be included. Excluding the data before the change time prevents previous states that are no longer relevant from affecting the forecast. As a result forecast accuracy may be improved.”; [0089]-[0091] – “At 720, the process selects time-series data that is more recent than the identified abnormal state change. In order to select the data, the process may compare the timestamp or other change time identifier against timestamps for the time-series data. Once the timestamp is matched to the timestamp of a sample data point in the time-series signal, data points in the time-series signal may be truncated. […] At 730, the process trains a forecasting model using the selected time-series data. […] At 740, the process generates a forecast using the trained forecasting model. […] These projections may be useful in a variety of applications, including, but not limited to, anomaly detection and capacity and planning operations.”).
Garvey is considered to be analogous art to the claimed invention because it is in the same field of predicting events using a model. Therefore, it would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal in view of Shashikant Rao to incorporate the teachings of Garvey such that training the model comprises identifying an event change point in the historical time-series data, such that training the model (including calculating an event time slot as taught by Neznal) and generating subsequent predictions using the model are based on the identified event change point. Identifying a change point in time-series data and excluding data from before the latest change point in training a prediction model ensures data, e.g., patterns that existed in the data, from before the change point and are no longer relevant do not impact the forecast, and thus improves the accuracy of the predictions made by the forecasting model (Garvey: [0086]).
Regarding claim 4, the combination of Neznal in view of Shashikant Rao and Garvey teaches the method of claim 1. Neznal further teaches wherein an event frequency comprises: one of daily, weekly, monthly quarterly, or yearly ([0027] – “commercial transaction features are compiled per hour, per day, or over other time periods during which commercial transactions are observer to be similar or have repeating patterns”; [0052] – “a time over which patterns are likely to repeat such as a day, a week, a month”).
Regarding claim 5, the combination of Neznal in view of Shashikant Rao and Garvey teaches the method of claim 4. Shashikant Rao further teaches wherein calculating the event frequency comprises: calculating a mean frequency for the event ([0101] – “the prediction system 1 4 0 generates a prediction based on scale optimization. As a specific illustrative example, time series of enormous scale can be predicted. […] In this example, the average monthly volume of time series transactions by account and/or category for a respective user may be 10, 15, or other average monthly volume based on varying time series”).
It would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal to incorporate the teachings of Shashikant Rao including calculating a mean frequency of the time series data (representing streams of events) and using different models based on the calculated event frequency. The methods of Shashikant Rao, including using different models for time series data with different sparsities (e.g., due to different event frequencies), allows for faster and more accurate predictions than what could be predicted by a single algorithm (Shashikant Rao: [0002], [0005], and [0039]).
Regarding claim 6, the combination of Neznal in view of Shashikant Rao and Garvey teaches the method of claim 1. Shashikant Rao further teaches wherein the first model of the plurality of models comprises one of: a seasonality model, a sequence model, a transformer model, a statistical model, or a rules-based model ([0037] – “It is to be understood that a model is selected based on other properties in addition to sparsity, including […] seasonality metrics”; [0039] – “a model is selected that utilizes a pattern of the time series, such as a seasonality pattern of the time series”; [0137] – “for non-sparse time series, a possible prediction algorithm is an auto-regressive approach including Autoregressive Integrated Moving Average (ARIMA)”).
It would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal to incorporate the teachings of Shashikant Rao such that the model used to generate a prediction of future events is selected from a plurality of models based on the calculated frequency of the at least one event and other properties of the time series data. Using a plurality of different models, including at least a seasonality model, for time series data with different sparsities (e.g., due to different event frequencies) and other features of the time series data allows for faster and more accurate predictions than what could be predicted by a single algorithm (Shashikant Rao: [0002], [0005], and [0039]).
Regarding claim 7, the combination of Neznal in view of Shashikant Rao and Garvey teaches the method of claim 1. Neznal further teaches wherein generating a prediction of one or more predicted future events based at least in part on the first model comprises:
estimating at least one event time within the calculated event time slot for the one or more predicted future events ([0053] – “At 1208, the high-dimensional time series windowed input is evaluated to generate an output of a predicted next element or elements in the series. At 1210, the predicted next element(s) output from the recurrent neural network are compared with the actual next element(s), and a difference metric is generated.”; [0032] – “the high-dimensional space further includes time-based information such as day of the week or hour of the day during which a transaction is conducted”. A next element of a high-dimensional time series is predicted (i.e., estimated), an element of the high-dimensional time series comprising information such as time based information of the transaction (“event”), e.g., a day (“event time”) of the week (“event time slot”), or hour (“event time”) of a day (“event time slot”) at which the transaction occurs.).
Regarding claim 8, the combination of Neznal in view of Shashikant Rao and Garvey teaches the method of claim 1. Shashikant Rao further teaches wherein the event frequency designations comprise one of: a frequent event, a moderate event, or a rare event ([0032] – “a first grouping may be low sparsity, a second grouping may be medium sparsity, and a third grouping may be high sparsity.”; [0040] – “time series data comprises an event stream. In one embodiment, an event is represented as a data point occurring in time, and an event stream is a stream of events”; [0131]-[0132] – “As illustrated in the first time series chart 710, a time series having a high sparsity property has fewer data points over time. In this specific illustrative example, the first time series chart 710 represents transfer of funds between accounts, which tends to be less frequent. […] As illustrated in the second time series chart 810, a time series having a low sparsity has more data points over time. In this specific illustrative example, the second time series chart 810 represents income of funds into accounts, which tends to be more frequent and denser.” A time series with a low sparsity grouping is analogous to a “frequent event” event frequency designation, a medium sparsity grouping is analogous to a “moderate event” event frequency designation, and a high sparsity grouping is analogous to a “rare event” event frequency designation.).
It would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal to incorporate the teachings of Shashikant Rao such that the model used to generate a prediction of future events is selected from a plurality of models based on the calculated frequency of the at least one event. Using different models for time series data with different sparsities (e.g., due to different event frequencies) allows for faster and more accurate predictions than what could be predicted by a single algorithm (Shashikant Rao: [0002], [0005], and [0039]).
Regarding claim 9, Neznal teaches a system ([0056] – “FIG. 13 is a computerized network commerce system comprising a recurrent neural network module, consistent with an example embodiment of the invention. FIG. 13 illustrates only one particular example of computing device 1300”), comprising:
a computer having a processor and a memory ([0057] – “computing device 1300 includes one or more processors 1302, memory 1304”); and
one or more code sets stored in the memory and executed by the processor, which, when executed, configure the processor ([0059] – “Processors 1302, in one example, are configured to implement functionality and/or process instructions for execution within computing device 1300. For example, processors 1302 may be capable of processing instructions stored in storage device 1012 or memory 1304”) to: perform the method of claim 1. Accordingly, claim 9 is rejected as being unpatentable over Neznal in view of Shashikant Rao and Garvey for the same reasons presented with respect to claim 1.
Claims 12-16 recite the same additional limitations as those recited in claims 4-8, applied to the system of claim 9. Accordingly, claims 12-16 are rejected as being unpatentable over Neznal in view of Shashikant Rao and Garvey for the same reasons presented with respect to claims 4-8.
Regarding claim 17, Neznal teaches a non-transitory computer-readable medium storing computer-program instructions that, when executed by one or more processors, cause the one or more processors to effectuate operations ([0059] – “processors 1302 may be capable of processing instructions stored in storage device 1012”; [0060] – “One or more storage devices 1312 may be configured to store information within computing device 1300 during operation. Storage device 1312, in some examples, is known as a computer-readable storage medium.”) comprising: the method of claim 1. Accordingly, claim 17 is rejected as being unpatentable over Neznal in view of Shashikant Rao and Garvey for the same reasons presented with respect to claim 1.
Claim 20 recites the same additional limitations as those recited in claim 8, applied to the non-transitory computer-readable medium of claim 17. Accordingly, claim 20 is rejected as being unpatentable over Neznal in view of Shashikant Rao and Garvey for the same reasons presented with respect to claim 8.
Claims 2-3, 10-11 and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Neznal in view of Shashikant Rao and Garvey as applied to claims 1, 9 and 17 above, and further in view of PETLADWALA et al. (U.S. Pub. No. 2021/0374031), hereinafter PETLADWALA.
Regarding claim 2, the combination of Neznal in view of Shashikant Rao and Garvey teaches the method of claim 1. Neznal teaches the method further comprising:
monitoring for the one or more predicted future events based on the calculated event frequency of the at least one event ([0052] – “the high-dimensional time series is windowed, such as by taking sequential overlapping groups of the time series, incremented by a time over which patterns are likely to repeat such as a day, a week, or a month, and provided to the recurrent neural network for training”; [0053] – “the trained recurrent neural network is implemented for monitoring live commercial transaction data at 1116”; [0054] – “commercial transactions are monitored at 1202 […] commercial transaction data is processed into a high-dimensional time series at 1204 and provided to a recurrent neural network at 1206, much as in the example of FIG. 11. At 1208, the high-dimensional time series windowed input is evaluated to generate an output of a predicted next element or elements in the series. At 1210, the predicted next element(s) output from the recurrent neural network are compared with the actual next element(s), and a difference metric is generated.”).
The combination of Neznal in view of Shashikant Rao and Garvey fails to expressly teach detecting when a given predicted future event of the one or more predicted future events does not occur.
However, PETLADWALA teaches monitoring for one or more events based on a calculated frequency of the one or more events ([0072]-[0074] – “The event monitoring apparatus 100 performs periodicity detection of time series data (Step S3). The event monitoring apparatus 100 performs subsequence time series clustering (Step S4). The event monitoring apparatus 100 performs event monitoring based on the result of the subsequence time series clustering (Step S5).”; [0067] – “Each cluster, that is, each column of the incidence matrix A may correspond to an event or state. The event analysis unit 160 calculates a periodicity for each column (column-wise periodicity) in the incidence matrix A.”; [0093] – “To find a periodicity, a frequency bin at which the maximum power spectral density (maximum peak) resides in the periodogram is selected.”; [0100] – “The period detection unit 124 obtains a time point N*T/kmax, the inverse number of which corresponds to the frequency at which the power spectrum density takes the maximum value. If a value of ACF at a time point (lag) N*T/kmax: ACF(N*T/kmax) takes a peak value (local maximum) (Yes branch of Step S33), the periodicity is set to N*T/kmax (Step S34).”); and detecting when a given predicted future event of the one or more predicted future events does not occur ([0004] – “The anomaly detection is also applied to identify a missed event”; [0066] – “The event analysis unit 160 may perform missing event detection by identifying the subsequence, occurrence of which is expected according to the periodicity of the subsequences belonging in common to one cluster, but not found”; [0075] – “When finding missed event (Yes branch of Step S6), the event monitoring apparatus 100 outputs notification of missing of event (Step S7).”; [0086] – “The missed event identification unit 163 checks whether or not the event occurs periodically (FIG. 5, Step S23). The missed event identification unit 163 identifies missing of a relevant event, when the element, which is expected to take a value of 1 in the incidence matrix A, takes a value of 0.”).
PETLADWALA is considered to be analogous art to the claimed invention because it is reasonably pertinent to the problem faced by the inventor of detecting missed events. Therefore, it would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal in view of Shashikant Rao and Garvey to not only monitor for the predicted future events but to also detect when a predicted future event does not occur as taught by PETLADWALA. Detecting missed events, e.g., in anomaly detection for monitoring time series, is useful for detecting failures and triggering analysis to find a reason for the failure (PETLADWALA: [0004]).
Regarding claim 3, the combination of Neznal in view of Shashikant Rao and Garvey and further in view of PETLADWALA teaches the method of claim 2. PETLADWALA teaches the method further comprising: initiating an alert when the given predicted future event of the one or more predicted future events does not occur ([0069] – “The output device 104 outputs a missed event. The output device 104 may include a display unit to output the missed event or a communication unit to transmit the missed event via a network to a terminal of a maintenance personnel.”; [0075] – “When finding missed event (Yes branch of Step S6), the event monitoring apparatus 100 outputs notification of missing of event (Step S7).”; [0087] – “When the missed event identification unit 163 identifies a missing event, the missed event identification unit 163 may notify the missing of event (alert) to the output device 104 (FIG. 5, Step S24). The output device 104 may output an occurrence of missed event on a display or notify alert by sound.”).
It would have been obvious to one of ordinary skill in the art to have modified the teachings of Neznal in view of Shashikant Rao and Garvey to not only monitor for the predicted future events but to also detect when a predicted future event does not occur and initiate and alert as taught by PETLADWALA. Detecting and providing notification of missed events, e.g., in anomaly detection for monitoring time series, is useful for detecting failures and triggering analysis to find a reason for the failure (PETLADWALA: [0004]).
Claims 10-11 recite the same additional limitations as those recited in claims 2-3, applied to the system of claim 9. Accordingly, claims 10-11 are rejected as being unpatentable over Neznal in view of Shashikant Rao and Garvey, and further in view of PETLADWALA for the same reasons presented with respect to claims 2-3.
Claims 18-19 recite the same additional limitations as those recited in claims 2-3, applied to the non-transitory computer-readable medium of claim 17. Accordingly, claims 18-19 are rejected as being unpatentable over Neznal in view of Shashikant Rao and Garvey, and further in view of PETLADWALA for the same reasons presented with respect to claims 2-3.
Response to Arguments
Applicant's arguments with respect to the rejections the claims 35 U.S.C. 101, see pages 6-17 of the Remarks filed 7/20/2026, have been fully considered but they are not persuasive.
Regarding Applicant’s argument that claim 1 recites “a specific ordered process for predicting repeatable events using event-frequency-based model selection, training the selected model using historical event data including event-change-point and event-time-slot processing, and generating predicted future events” and that the limitations of claim 1 “define a particular technical workflow for learning event patterns and generating predicted future events, not a mental process.” (see page 7 of the Remarks):
Examiner disagrees. As identified in the rejection of claim 1, the claims recite a series of steps which constitute an additional element of insignificant extra-solution activity of mere data gathering and a series of steps which can be performed in the human mind, through observation, evaluation, judgment and opinion, with the aid of pen and paper. The broadest reasonable interpretation of the steps which the Applicant has identified as defining “a particular technical workflow for learning event patterns and generating predicted future events” has been identified as a mental process by the Examiner, and the Applicant has not provided evidence in this argument to the contrary. Thus, when considering the claim as a whole, the claim amounts to gathering data and using the gathered data in steps which can performed in the human mind, through observation, evaluation, judgment and opinion, with the aid of pen and paper, which are judicial exceptions, and the claim as a whole does not amount to significantly more than the recited judicial exception.
Regarding Applicant’s argument that the Specification discloses a technical problem, and that the claim provides a technical solution to the disclosed problem (see pages 7-9):
Examiner disagrees. The Applicant has pointed to the steps of claim 1 including “calculating event frequency, selecting a model based on the calculated event frequency, calculating an event time slot based on an event change point, and generating predicted future events”, all of which have been identified as encompassing a mental process by the Examiner. Regarding evaluating improvements to technical fields, the MPEP states “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. See the discussion of Diamond v. Diehr, 450 U.S. 175, 187 and 191-92, 209 USPQ 1, 10 (1981)) in subsection II, below. In addition, the improvement can be provided by the additional element(s) in combination with the recited judicial exception.” (see MPEP 2106.05(a)). The only additional element recited by the claim is “receiving an event identifier and historical data for at least one event” which amounts to insignificant extra-solution activity of mere data gathering. When considering the claim as a whole, the combination of this additional element with the recited judicial exceptions results in gathering data that is subsequently used in performing the recited judicial exceptions. This combination is not indicative of providing a technological improvement.
Further, the alleged improvements are not reflected in the claims:
The alleged improvement of “automating the monitoring of important events" is provided by merely automating a manual monitoring process, e.g., by using a generic computer to perform steps which, as recited in claim 1 and indicated in the rejection of claim 1, can be performed in the human mind, through observation, evaluation, judgment and opinion, with the aid of pen and paper. Mere automation of a manual process is not sufficient to show a technical improvement (see MPEP 2106.05). Further, claim 1 does not even require use of a computer, and as such, cannot reflect the alleged improvement. With respect to the generic computing components recited in claims 9 and 17 which have been identified as mere instructions to apply the exception, “To show that the involvement of a computer assists in improving the technology, the claims must recite the details regarding how a computer aids the method, the extent to which the computer aids the method, or the significance of a computer to the performance of the method. Merely adding generic computer components to perform the method is not sufficient.” (see MPEP 2106.05(a)).
The alleged improvement of "reducing false positives" is provided by removing inaccurate or outdated data from the historical data, e.g., by removing data from before the latest change point, such that the predictions are not generated based on the inaccurate data, which is not reflected in the claim (see Specification [0009] and [0065]-[0066]). It also teaches this unreliable information about event patterns may be due to the factor “event information is not shared with the monitoring entity” (see Specification [0009]); however, the only limitation in the claims which reflects this issue being overcome is “receiving an event identifier and historical data for at least one event”, which as identified in the rejection of claim 1, amounts to insignificant extra-solution activity of mere data gathering and is not indicative of integration into a practical application (i.e., providing a technological improvement), nor is it enough to amount to significantly more than the recited judicial exception (see MPEP 2106.04(d) and 2106.05(a), (d) and (g)).
The alleged improvement of “providing early indicators of potential problems in business or technology process” is not reflected in the claim, as generating a prediction does not require that a problem has been predicted nor that an indicator of such a problem is provided. For clarity of the record, claim 3 is the closest to reflecting this alleged improvement; however, “initiating an alert” when the predicted event does not occur is not the same as actually providing an alert, and the limitation is recited in such a way that the human mind, through observation, evaluation, judgement and opinion, with the aid of pen and paper, would be capable of performing the step.
The alleged improvement of "enabling integration of prediction models with other tools having events" is not reflected in the claims. The claim does not require any sort of integration of the steps of the method with other “tools”, e.g., receiving data from other tools or providing predictions back to the tools; rather it merely requires that data for at least one event is received and a prediction is generated based at least in part on one of the models.
Regarding Applicant’s arguments with respect to Step 2A, Prong One (see pages 9-12):
Applicant alleges “A mental process involves concepts that can practically be performed in the human mind, such as observations, evaluations, judgments, or opinions. Claim 1 does not merely recite observing that a predicted event did not occur. Instead, claim 1 requires calculating an event frequency, identifying a first model of a plurality of models based on the calculated event frequency, training the first model based on historical event data, identifying an event change point in the historical data, calculating an event time slot based on the event change point, and generating predicted future events based on the trained model. These operations, considered together, are not practically performed in the human mind.”
Examiner disagrees. As stated in MPEP 2106.04(a)(III), “claims do recite a mental process when they contain limitations that can practically be performed in the human mind, including for example, observations, evaluations, judgments, and opinions.” A mental process is not limited to an observation.
Claim 1 requires “calculating an event frequency” (an observation, evaluation or judgment), “identifying a first model of a plurality of models based on the calculated event frequency” (a judgment), “training the first model based on historical event data” (an evaluation), “identifying an event change point in the historical data” (an observation), “calculating an event time slot based on the event change point” (an observation, evaluation or judgment), and “generating predicted future events based on the trained model” (an evaluation or judgment). Thus, these limitations are all reciting steps which can performed in the human mind through observations, evaluations, judgments and opinions, as explained in the rejection of claim 1.
Applicant’s arguments that “The claim-recited calculations and model training therefore operate within a computer-implemented event-monitoring architecture that processes historical event data, model outputs, event time slots, and predicted future events” are not persuasive. Claim 1 entirely fails to recite any computer or computer-implemented system/components. Thus, there is no indication in the claim that a computer is being used in performance of any of the recited limitations.
The Examiner would like note the recitation of a computer in claims 9 and 17 amounts to no more than mere instructions to apply the exception, as they are merely reciting a generic computer is used to implement the method of claim 1.
Applicant’s argument “The claimed event-frequency calculation is likewise not merely a mental judgment” is not persuasive. The Applicant relies on specific steps/calculations described in the Specification in their arguments which are not required by the limitations of the claims. Further, the argued steps could still be performed in the human mind through observation, evaluation, judgment or opinion (calculating mean values, removing outliers, calculating an average, mapping frequencies to categories, etc.), and could additionally be considered a mathematical calculation, which is also a judicial exception of an abstract idea (see 2106.04(a)).
Applicant’s argument “Nor does claim 1 merely recite use of a generic model. Claim 1 requires identifying the first model based on the calculated event frequency, and further requires that different models are associated with different event-frequency designations” is not persuasive. Claim 1 does not require a specific type of model, only that the model is “identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations”. Requiring that different models are “associated with” different event-frequency designations could amount to a plurality of different models being labeled with a different frequency designation (e.g., one model is labeled as being used for daily events, another model is labeled as being used for weekly events). Based on a calculated event frequency and the labels corresponding to the models, a person could judge (“identify”) which model should be used. Further, person of ordinary skill in the art, e.g., a computer programmer, would be capable of selecting (“identifying”) an appropriate model (i.e., particular logic such as a set of rules, e.g., “a rule-based model”, see Specification: [0062]), from a plurality of known types of models, to use based on the event frequency though judgment.
Applicant’s argument “The claim-recited event-change-point and event-time-slot operations also are not mental processes. […] These limitations are directed to technical treatment of historical event data to improve prediction accuracy, not to a mental process” is not persuasive. The Applicant is adding in requirements from the Specification which are not recited by the limitations claims. The claims merely require at least one event change point in the historical data is identified (without limitation as to how it is identified), and that an event time slot is calculated based on the at least one event change point, both of which can be performed in the human mind as identified with respect to previous arguments and in the rejection of claim 1. There is no requirement in the claim that the at least one change point being identified and used in calculation of the event is the “latest change point”, nor that the event time slot is calculated based on the historical data after the latest change point, as would be required to provide the technical benefit being argued (minimization of noise and improving accuracy). For example, the claims as currently recited encompass the embodiment in which the latest change point is identified, and the training data prior to the latest change point is used in calculating the event time slot, which would result in a more inaccurate prediction as it does not account for the new pattern of the data after the change point.
Applicant’s argument “The specification further describes the claimed event-time-slot calculation as a concrete data transformation. […] These operations are data-structure generation and labeling operations for training and prediction, not practical mental steps” is not persuasive. The claims fail to recite any detail regarding how the step “calculating an event time slot based on the at least one event change point in the historical data” is performed; thus, it is not limited to the explicit teachings of the Specification as argued by the Applicant, and may include a mental process as identified with respect to previous arguments and in the rejection of claim 1.
Applicant’s arguments “The Office Action's mental-process characterization also abstracts away the claimed generation of future-event predictions based on the selected model. […] These operations generate predicted event slots and event times from historical event data and selected models. They are not mental observations” are not persuasive. The Applicant is arguing limitations read in from the Specification (and not required by the limitations of the claim) provide evidence that the limitation recited by the claim does not encompass a mental process. All that claim 1 requires is “generating a prediction of one or more predicted future events based at least in part on the first model”, where “the first model is identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations”. As explained in the rejection of claim 1, as person is capable of performing such steps through observation, evaluation, judgment and opinion. The claims do not require the model itself generates the prediction, nor that the model is executed/performed by a computer, but merely require that a prediction is generated based at least in part on the selected model, e.g., on an output of the model.
Regarding Applicant’s arguments with respect to Step 2A, Prong Two (see pages 12-14):
Applicant’s argument “Claim 1 applies any alleged event-pattern analysis in a specific technological context: automated prediction of repeatable technical and service events and detection of missed predicted events […] The specification's disclosed implementation confirms the practical application. […] Thus, the claimed event-prediction workflow is applied to monitoring technical
and service events that occur in networked computing environments” is not persuasive. The claims broadly recite “event”, which is not limited to the technical and service events described by the Specification and being argued by the Applicant.
Applicant’s argument “The specification further describes the claimed improvement in terms of monitoring reliability. […] This is a practical application that improves event-monitoring and missed-event-detection systems by reducing false positives and improving reliability” is not persuasive. Claim 1 is reciting a method which can be performed entirely in the human mind, through observation, evaluation, judgment and opinion. Please note the step “receiving an event identifier and historical data for at least one event;”, although identified as insignificant extra-solution activity, may also be performed in the human mind through observation and encompasses a judicial exception. See the rejection of claim 1 above indicating how the remaining steps are performed in the human mind. “Because a judicial exception alone is not eligible subject matter, if there are no additional claim elements besides the judicial exception, or if the additional claim elements merely recite another judicial exception, that is insufficient to integrate the judicial exception into a practical application. However, the way in which the additional elements use or interact with the exception may integrate it into a practical application.” (see MPEP 2106.04(d)(III). Additionally, “in order for a method claim to improve computer functionality, the broadest reasonable interpretation of the claim must be limited to computer implementation. That is, a claim whose entire scope can be performed mentally, cannot be said to improve computer technology.” See MPEP 2106.05(a)(I). The method of claim 1 is not limited to a computer implementation, and therefore cannot be said to improve computer-implemented monitoring systems as argued.
Applicant’s arguments that the claims provide the alleged improvements of “improv[ing] event-monitoring and missed-event-detection systems by reducing false positives and improving reliability” are not persuasive. The Specification discloses those improvements are provided by removing inaccurate or outdated data from the historical data, e.g., by removing data from before the latest change point, such that the predictions are not generated based on the inaccurate data, which is not reflected in the claim (see Specification [0009] and [0065]-[0066]). It also teaches this unreliable information about event patterns may be due to the factor “event information is not shared with the monitoring entity” (see Specification [0009]); however, the only limitation in the claims which reflects this issue being overcome is “receiving an event identifier and historical data for at least one event”, which as identified in the rejection of claim 1, amounts to insignificant extra-solution activity of mere data gathering and is not indicative of integration into a practical application (i.e., providing a technological improvement), nor is it enough to amount to significantly more than the recited judicial exception (see MPEP 2106.04(d) and 2106.05(a), (d) and (g)).
Applicant’s argument that “Here, the claims are directed to event prediction and missed-event detection, which the specification identifies as improving legacy monitoring systems by "reducing false positives" and "providing early indicators of potential problems in business or technology processes." Spec. [0030]. As in the eligible AI-related examples, the claimed operations are not the end in themselves; they are integrated into a practical monitoring architecture that predicts future events and detects missed events in a technical/service framework” is not persuasive. As explained previously, the claims do not reflect the improvement of “reducing false positives” nor “providing early indicators of potential problems in business or technology processes” as key steps disclosed by the Specification as providing such benefits are omitted from the claims. Further claim 1 does not reflect detecting missed events, nor do any of the limitations of claim 1 require the method occurring in “a technical/service framework”.
Applicant’s argument that “The claimed architecture also imposes meaningful limits on any alleged abstract idea” is not persuasive. As stated in the MPEP, an alleged improvement indicating integration of a judicial exception into a practical application cannot be provided by the judicial exception, but rather must be provided by the additional elements or combination of the additional elements with the judicial exceptions (see MPEP 21.06.04(d) and 2106.05(a)). The only additional element identified in claim 1 is “receiving an event identifier and historical data for at least one event”, which when considered alone or in combination with the recited judicial exception, amounts to insignificant extra-solution activity of mere data gathering, where the data gathered is then recited as being used in performing the judicial exception. This additional element of insignificant extra-solution activity does not meaningfully limit the judicial exceptions recited by the claims (see MPEP 2106.04(d) and 2106.05(g)).
Applicant does not present specific arguments as to how the limitations recited by the dependent claims integrate the recited judicial exception into a practical application. As identified in the rejections of the claims, all limitations of the dependent claims have been identified as reciting limitations which can be performed in the human mind, through observation, evaluation, judgment, and opinion, with the aid of pen and paper. As stated in MPEP 2106.04(d), “if the additional claim elements merely recite another judicial exception, that is insufficient to integrate the judicial exception into a practical application.”
Regarding Applicant’s arguments with respect to Step 2B (see pages 14-16):
Applicant’s arguments that the ordered combination of elements presented in the claims provide the benefits disclosed by the Specification (e.g., generating more reliable predicted events and reducing false positives to improve the operation of monitoring systems) are not persuasive.
The Examiner has already identified that the claims do not reflect the improvements disclosed in the Specification (see Examiner’s response to arguments on pages 23-25 of the instant Office Action with reference to the arguments presented on pages 7-9 of the Remarks). Further, as stated above, regarding evaluating improvements to technical fields, the MPEP states “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. See the discussion of Diamond v. Diehr, 450 U.S. 175, 187 and 191-92, 209 USPQ 1, 10 (1981)) in subsection II, below. In addition, the improvement can be provided by the additional element(s) in combination with the recited judicial exception.” (see MPEP 2106.05(a)). The only additional element recited by the claim is “receiving an event identifier and historical data for at least one event” which amounts to insignificant extra-solution activity of mere data gathering. When considering the claim as a whole, the combination of this additional element with the recited judicial exceptions results in gathering data that is subsequently used in performing the recited judicial exceptions. This combination is not indicative of providing a technological improvement nor amounting to significantly more than the recited judicial exception.
Applicant’s arguments that “The specification also describes technical details of how the ordered combination improves prediction reliability” are not persuasive as the details cited which provide improved prediction reliability are not included in the claims. Specifically, the claims do not require selecting training data occurring after the latest change point, as described in [0065]-[0066] of the Specification, such that the inaccurate data from before a change point is not involved in the subsequent steps disclosed in [0067]-[0079].
Applicant’s arguments that “The Office Action has not shown that this ordered architecture was well-understood, routine, and conventional” are not persuasive. Only the additional elements and combination of additional elements must be well-understood, routine and conventional (see MPEP 2106.05(a)(B) “Examples Of How Courts Conduct The Search For An Inventive Concept […] The Court considered the additional elements "as an ordered combination," and determined that "the computer components … ‘[a]dd nothing … that is not already present when the steps are considered separately’"). Further, with respect to BASCOM Global Internet Services, Inc. v. AT&T Mobility LLC, 827 F.3d 1341 (Fed. Cir. 2016), “BASCOM provides another example of how courts conduct the significantly more analysis, and of the critical importance of considering the additional elements in combination. […] The district court should have considered the additional elements in combination, because the "inventive concept inquiry requires more than recognizing that each claim element, by itself, was known in the art" (see MPEP 2106.05(a)). The additional elements recited in the claims are “receiving an event identifier and historical data for at least one event” as recited in claim 1; “A system, comprising: a computer having a processor and a memory; and one or more code sets stored in the memory and executed by the processor, which, when executed, configure the processor to: receive an event identifier and historical data for at least one event;” as recited in claim 9; and “A non-transitory computer-readable medium storing computer-program instructions that, when executed by one or more processors, cause the one or more processors to effectuate operations comprising: receiving an event identifier and historical data for at least one event;” as recited in claim 17. As presented in the rejections above, “receiving an event identifier and historical data for at least one event” is insignificant extra-solution activity of mere data gathering (see MPEP 2106.05(g)), and also falls within the well-understood, routine and convention activity of receiving data over a network as identified by the courts (see MPEP 2106.05(d)). The other additional elements ““A system, comprising: a computer having a processor and a memory; and one or more code sets stored in the memory and executed by the processor, which, when executed, configure the processor to:” and “A non-transitory computer-readable medium storing computer-program instructions that, when executed by one or more processors, cause the one or more processors to effectuate operations comprising:” are reciting generic (i.e., well-known) computing components are implementing the method of claim 1, which is mere instructions to apply the exception (see MPEP 2106.05(f)). None of the combinations of the additional elements in claim 1, 9, or 17 amount to significantly more than the recited judicial exceptions, as they are merely reciting a generic computer is used to perform the judicial exceptions, and that data used in performing the judicial exceptions is gathered. Further a generic computer receiving data, e.g., over a network, is well-known in the art (see MPEP 2106.05(d) and (f)).
Applicant’s arguments “Nor is this a case in which a claim merely applies generic machine learning to a new field of use. The claims do not simply recite "use a model to predict events." Instead, the claims recite selecting a model from a plurality of models based on a calculated event frequency and associating different models with different event-frequency designations. The specification explains why this matters: frequent events may have enough data for separate seasonality models; moderate events may be handled using a sequence model trained across weekly events; and rare events may lack sufficient data for machine-learning training and may therefore be handled using a rules-based model” are not persuasive. Merely linking a plurality of different “models” (which are not limited to “machine learning” models, see Specification: [0062] – “rather than implementing a machine learning model, in some embodiments, a rule-based model or system may be implemented”) to different event frequency designations, i.e., “wherein different models are associated with different event frequency designations”, as recited by the claims does not teach a using a particular type of model that best, or even appropriately, fits the frequency of the event data.
Applicant’s argument “The claims further recite event-change-point identification and event-time-slot calculation, which the specification ties to reducing inaccuracies from changed event schedules” is not persuasive. Reducing inaccuracies from changed event schedules is provided by excluding data prior to the latest event change point from the training data as disclosed in paragraphs [0064]-[0067] of the Specification. This step to provide the improvement is not present in the claims.
No additional arguments are presented with respect to claims 2-20. Accordingly, the Examiner’s response to the arguments presented with respect to claim 1 is applicable to claims 2-20 as well.
Applicant's arguments with respect to the rejections the claims 35 U.S.C. 103, see pages 17-24 of the Remarks filed 7/20/2026, have been fully considered but they are not persuasive.
Regarding Applicant’s allegation the Neznal does not teach the claimed “calculating an event frequency of the at least one event” (see page 18 of the Remarks):
Examiner disagrees. The Examiner acknowledges the Specification defines “event frequency” as “frequency of the occurrence of an event ( e.g., "repeatable events"), e.g. , 1 hour, 2 hours, daily, weekly, monthly, quarterly, and/or yearly, etc.” (see Specification: paragraph [0032]). Thus, the claims reciting “calculating an event frequency of the at least one event” is understood to mean calculating a frequency of occurrence of the at least one event, i.e., calculating the frequency at which the at least one event repeats. The claims do require a particular manner in which the event frequency of the at least one event is calculated as is being argued by the Applicant (e.g., by the “multi-step calculation methodology disclosed in the Specification). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993).
The plain meaning of “calculate” is not limited to performing a mathematical computation or series of steps, but rather includes making a judgement or determination (see the Merriam Webster definition of “calculate”, specifically the definition 1b: “to reckon by exercise of practical judgment”, cited below as pertinent prior art).
Thus, the teachings of Neznal cited with respect to claim 1, teach “calculating an event frequency of the at least one event”, as it implicitly teaches judging/determining the frequency at which the time series data of the transactions (“at least one event”) repeat (e.g., daily, weekly, monthly, etc.) when pre-processing the time series data of the transactions into windowed time series data (see [0027] – “commercial transaction features are compiled per hour, per day, or over other time periods during which commercial transactions are observed to be similar or have repeating patterns."; [0041]-[0042]; [0052] – “At 1106, the high-dimensional time series is windowed, such as by taking sequential overlapping groups of the time series, incremented by a time over which patterns are likely to repeat such as a day, a week, or a month, and provided to the recurrent neural network for training.”).
Regarding Applicant’s allegation that Shashikant Rao does not teach “identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations” (see pages 19-20 of the Remarks):
Examiner disagrees. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). The combination of Neznal in view of Shashikant Rao is relied upon to teach the limitations “calculating an event frequency of the at least one event; identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations”, where Neznal is relied upon to teach “calculating an event frequency of the at least one event” (see at least Neznal: [0027] – time series data of identified transactions are windowed based on a calculated event frequency the transactions (e.g., hourly, daily, etc.) and Shashikant Rao is relied upon to teach “identifying a first model of a plurality of models, wherein the first model is identified based on [an] event frequency of the at least one event, and wherein different models are associated with different event frequency designations” (see at least Shashikant Rao: [0028]-[0029] – a sparsity level of a time series is related directly to the frequency of the one or more events, e.g., if events occur/repeat every two weeks, once a quarter, etc.; [0032] – “the time series data sets have varying levels of sparsities. As a specific illustrative example, a time series is dense when transactions are generated hourly and is sparse when transactions are generated weekly. In one embodiment, time series are grouped by sparsity levels as discussed further below with respect to FIGS. 7 and 8. In one embodiment, an appropriate model is chosen based on a sparsity level grouping. In one embodiment, sparsity level groupings define subsets of time series. As a specific illustrative example, a first grouping may be low sparsity, a second grouping may be medium sparsity, and a third grouping may be high sparsity.”; [0040]; [0131]-[0132] – a time series with a high sparsity property (i.e. high-sparsity grouping) has fewer data points over time, due to the event frequency of the time series of transactions being less frequent; likewise, transactions (“at least one event”) that are more frequent produce a time series with lower sparsity.).
Specifically, in the teachings of Shashikant Rao, since 1) a sparsity level of time series data is directly related to/based on the frequency of the transactions (“at least one event”) in the time series, and 2) a model is identified/selected based on the sparsity level, then 3) the model is therefore identified/selected based on the frequency of the transactions of the time series data (“event frequency”), e.g. via the sparsity level.
Thus, when the teachings of Shashikant Rao are applied to Neznal, the event frequency (e.g., hourly, weekly, etc.) of the time series data based on which the first model is identified/selected as taught by Shashikant Rao may be the event frequency (e.g., hourly, weekly, etc.) calculated for the time series data in Neznal.
Attacking Shashikant Rao individually is not persuasive.
Applicant’s arguments that “that Rao's sparsity groupings do not teach the claimed event-frequency designations” are not persuasive. Shashikant Rao teaches the sparsity of a time series is directly related to the frequency of the at least one event in the time series data (see cited portions of Shashikant Rao: [0028]-[0029] – “a time series data set of payroll transactions may be moderately sparse due to a moderate volume of paychecks that are provided to employees every two weeks. As a specific illustrative example, a time series data set of government regulation transactions may be severely sparse due to a low volume of payments to regulatory entities that occur once a quarter. It is to be understood that the infrequency of transactions of a time series can impact the sparsity of the time series.”; [0032] – “the time series data sets have varying levels of sparsities. As a specific illustrative example, a time series is dense when transactions are generated hourly and is sparse when transactions are generated weekly. In one embodiment, time series are grouped by sparsity levels as discussed further below with respect to FIGS. 7 and 8. In one embodiment, an appropriate model is chosen based on a sparsity level grouping. In one embodiment, sparsity level groupings define subsets of time series. As a specific illustrative example, a first grouping may be low sparsity, a second grouping may be medium sparsity, and a third grouping may be high sparsity.”; [0131]-[0132] – “a first time series chart 710 associated with transfers, which depicts high sparsity. As illustrated in the first time series chart 710, a time series having a high sparsity property has fewer data points over time. In this specific illustrative example, the first time series chart 710 represents transfer of funds between accounts, which tends to be less frequent. […] a second time series chart 810 associated with income, which depicts low sparsity. As illustrated in the second time series chart 810, a time series having a low sparsity has more data points over time. In this specific illustrative example, the second time series chart 810 represents income of funds into accounts, which tends to be more frequent and denser.”). A time series of transactions with a lower frequency, e.g., weekly, has a high sparsity property (the “sparsity level grouping” is high sparsity), and a time series of transactions with a greater frequency, e.g., hourly, is dense and has a low sparsity property (the “sparsity level grouping” is low sparsity).
Thus, the sparsity level grouping for a time series is directly related to/based on the event frequency of the transactions of the time series, and can be considered “event frequency designations”. Since the models are selected based on the corresponding sparsity levels of the time series (see Shashikant Rao: [0032]), the different models are “associated with different” sparsity levels, i.e., “event frequency designations”).
As explained previously, the combination of Neznal in view of Shashikant Rao teaches the limitations ““calculating an event frequency of the at least one event; identifying a first model of a plurality of models, wherein the first model is identified based on the calculated event frequency of the at least one event, and wherein different models are associated with different event frequency designations”. No specific additional arguments are presented with respect to the combination of Neznal in view of Shashikant Rao.
Regarding Applicant’s arguments with respect to "training the first model based on the historical data for the at least one event, wherein training the first model based on the historical data for the at least one event further comprises: identifying at least one event change point in the historical data; and calculating an event time slot based on the at least one event change point in the historical data" (see pages 20-23):
Applicant argues the teachings of Garvey “may relate generally to selecting post-change data for training a forecasting model. They do not teach calculating an event time slot based on an event change point” (see page 21 of the Remarks). The Examiner would like to point out that Garvey along is not relied upon to teach “calculating an event time slot based on the at least one event change point in the historical data” as recited in claim 1, but rather the combination of Neznal in view of Garvey is relied upon to teach the argued limitations. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
In response to applicant's argument that the references fail to show certain features of the invention (see pages 20-21 of the Remarks), it is noted that the features upon which applicant relies (i.e., “the claimed event time slot is not merely any historical-data window. It is a time interval generated for event prediction based on the event frequency and calculated using historical event data after an event change point”) are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993). Claim 1 only recites training the first model based on the historical data for the at least one event further comprises: “calculating an event time slot based on the at least one event change point in the historical data”. There is no requirement recited in the claim that it is “generated for event prediction based on event frequency” nor or that it is calculated “using event data after an event change point” as is being argued.
Regarding Applicant’s arguments with respect to Neznal, (see page 21 of the Remarks – “That input window is not an event time slot calculated based on an event change point. It is not generated from historical event data after an event change point, is not labeled to indicate event presence or absence within the slot, and is not used as the claimed event time slot for generating predicted future events”, and “Neznal's windowed input sequences do not teach the claimed relationship between the event change point and the calculated event time slot.”), Applicant is once again attacking Neznal individually when the combination of Neznal in view of Garvey is relied upon to teach the limitation being argued. Additionally, Applicant is once again arguing the references do not teach features which are not recited by claim 1 (that the event time slot “is not generated from historical event data after an event change point, is not labeled to indicate event presence or absence within the slot, and is not used as the claimed event time slot for generating predicted future events”). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. Accordingly, these arguments are not persuasive.
Applicant generally alleges the combination of Neznal in view of Shashikant Rao and Garvey does not teach the claimed training limitation.
Examiner disagrees. Neznal teaches “training the first model based on the historical data for the at least one event ([0028]), wherein training the first model based on the historical data for the at least one event further comprises: […] calculating an event time slot based on […] the historical data (see [0041]-[0042] and [0052] – training the model comprises predicting the next data element to be observed in the time series based on the historical data (the windowed time series data), e.g., for comparison against an actual next data element; [0032] – "high-dimensional space further includes time-based information such as day of the week or hour of the day during which a transaction is conducted”. As described in [0032], a data element of the time series includes time-based information such as a day of the week or hour of the day during which a transaction occurs; thus predicting a next data element of the time series would involve predicting (“calculating”) the time-based information for the data element, such as a day of the week or hour of the day during which the predicted transaction occurs (“an event time slot”). Please additionally see paragraphs [0034]-[0035] of Neznal, which provide further support.).
Garvey teaches training a prediction model comprises: identifying at least one event change point in the historical data (see portions of [0034]; [0080]; [0088]; and [0053]); and calculating an event time slot based on the at least one event change point in the historical data (see cited portions of [0086]; [0089]-[0091] which teach excluding data before an identified change point from being included in the historical data used for training a model to generate a forecast (i.e., predict future events) of a time series). Please note Garvey teaches a time series comprises a series of data values (“events”) over time, where the data values have a corresponding timestamp (see Garvey: [0038] and [0058]); thus, generating a forecast of a time series involves projecting future data values and their corresponding timestamps at which they occur (e.g., see Garvey: Figs. 8A-8B which depict plots of a forecast of future values over time).
Thus, when applying the teachings of Garvey to the teachings of Neznal, one would modify training the model based on historical data of the at least one event as taught by Neznal to include identifying at least one change point in the historical data as taught by Garvey, and would modify the historical data used in training the model as taught by Neznal to exclude data after the at least one identified change point as taught by Garvey, as doing so would provide the benefit of improving accuracy of the predictions made by the model (see Garvey: [0086]).
Thus, training the model, including predicting a next element of a time series (including its event time slot) based on the historical data as taught by Neznal, where the historical data used for training the model excludes data from prior to a change point in the time series data as taught by Garvey, is teaching “calculating the event time slot based on the at least on event change point in the historical data” as recited by claim 1.
No additional arguments are presented with respect to claims 2-20. Accordingly, the Examiner’s response to the arguments presented with respect to claim 1 is applicable to claims 2-20 as well.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Merriam-Webster (NPL Document “calculate” on attached PTO-892) teaches the definition of “calculate” includes ““to reckon by exercise of practical judgment” (see page 3).
Chanda et al. (U.S. Pub. No. 2022/0114074) teaches techniques for detecting anomalous values in data streams using forecasting models, where a data stream comprises a plurality of data values each corresponding to a time interval, where models are selected based on the time interval and each of the models has a different periodicity/sampling frequency, e.g., if the sampling frequency is once per day and the new data value corresponds to the first Monday of the month, a weekly model that corresponds to the Monday of each week (i.e., the Monday model) and a monthly model that corresponds to the first Monday of each month (i.e., the first monthly Monday model) may be selected (see Abstract, [0025], and [0072]).
Jermann et al. (U.S. Pub. No. 2020/0241490) teaches an event prediction model for a given type of event may output a prediction including an “event window” (see [0005]).
STANEVICH et al. (U.S. Pub. No. 2022/0414495) teaches predicting at a first machine learning model, a frequency of transactions over a given time period using historical frequency data from a prior time period, where the first machine learning model applies Poisson regression for characterizing distribution of the historical frequency data (see [0006] and [0009]).
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JENNIFER MARIE GUTMAN whose telephone number is (703)756-1572. The examiner can normally be reached M-F: 8:00 am - 4:00 pm.
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, Kevin Young can be reached at 571-270-3180. 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.
/JENNIFER MARIE GUTMAN/Examiner, Art Unit 2194 /KEVIN L YOUNG/Supervisory Patent Examiner, Art Unit 2194
