DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
This action is responsive to the Amendment filed on 03/27/2026. Claims 1-20 are pending in the case. This action is Final.
Applicant Response
In Applicant’s response dated 03/27/2026, Applicant amended Claims 1, 11 and 18 and argued against all objections and rejections previously set forth in the Office Action dated 01/05/2026.
Examiner Comments
4. In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
Claim Rejections - 35 USC § 101
5. 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 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1
According to the first part of the analysis, in the instant case, claims 1-10 are directed to a computer-implemented method, claims 11-17 are directed to a computer program product, and claim 18-20 is directed to a computer system. Thus, each of the claims falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Regarding Claim 1, 11 and 18,
At step 2A, prong 1, Does the claim recite a judicial exception?
Independent claims 1, 11, and 18, the steps of:
generating a feature vector by applying a feature engineering pipeline to raw data from a data warehouse, the feature engineering pipeline comprising cleaning the raw data by removing outliers and interpolating missing data to produce cleaned data and augmenting the cleaned data, the feature vector comprising a set of raw features including context-specific data associated with a time period and a location identifier and a plurality of lag features generated using a sliding-window approach over past time periods, the plurality of lag features including a current value of a selected feature in the set of raw features and one or more historical values of the selected feature (This step for generating a feature vector is a mathematical calculations and predictive modeling operations which fall within the “Mathematical Concepts” grouping of abstract ideals.)
inputting the feature vector into a plurality of base models of different model types, the plurality of base models including at least a decision tree-based model comprising a gradient-boosting model and a recurrent neural network model comprising a long short-term memory network having an LSTM portion that receives the plurality of lag features and generates a first prediction, wherein the first prediction is concatenated with the feature vector and input into one or more dense layers to generate a neural network prediction and outputting a plurality of predictions, each prediction in the plurality of predictions representing future values of the selected feature, each base model processing the same feature vector to generate its respective prediction; (This step for inputting a feature vector is a mathematical calculations and predictive modeling operations which fall within the “Mathematical Concepts” grouping of abstract ideals.)
predicting a future value of the selected feature by inputting the plurality of predictions into a meta-model comprising a linear regression model that applies a learned weighting to each of the plurality of predictions, the meta-model applying a learned weighting to each of the plurality of predictions to generate an ensemble prediction with improved accuracy compared to any individual base model prediction by combining predictions from the different model types to reduce prediction variance and improve generalization performance across different data distributions (This step for predicting a future value is a mathematical calculations and predictive modeling operations which fall within the “Mathematical Concepts” grouping of abstract ideals.)
The claim recites judicial exception because it recites mathematical concepts through feature engineering, machine learning prediction, regression weighting and ensemble forecasting which falls within the “mathematical concepts” groupings of abstract ideas. Accordingly, the claims recite an abstract idea.
Step 2A prong 2: Does the claim recite additional elements? Do those additional elements, individually and in combination, integrate the judicial exception into a practical application?
a non-transitory computer-readable storage medium for tangibly storing computer program instructions capable for execution by a computer processor, (claim 11) and a system comprising: a processor (claim 18): (These steps describe mere instructions to apply the exception using generic computer components - see MPEP 2106.05(f)).
controlling a computing system based on the predicted future value of the selected feature, wherein the computing system includes at least one of a fraud detection system, an inventory management system, or a resource allocation system wherein controlling comprises automatically adjusting operational parameters of the computing system to respond to the predicted future value (This step additional element that Merly apply the mathematical prediction in a generic computer environment)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are insignificant extra solution activity in combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
Step 2B: Do the additional elements, considered individually and in combination, amount to significantly more than the judicial exception?
No, As shown above with respect to integration of the abstract idea into a practical application, the additional element of:
controlling a computing system based on the predicted future value of the selected feature, wherein the computing system includes at least one of a fraud detection system, an inventory management system, or a resource allocation system wherein controlling comprises automatically adjusting operational parameters of the computing system to respond to the predicted future value (This step additional element that Merly apply the mathematical prediction in a generic computer environment)
The additional elements, alone and in combination, fail to integrate the abstract idea into a practical application. Thus, the claims are not patent eligible.
The dependent claims respectively recite a judicial exception in limitations of: “wherein iteratively creating the one or more additional sets of ANNs, comprises: generating an intrinsically augmented feature, the intrinsically augmented feature comprising one or more of a synthetic date feature, a historical feature, and an aggregate feature..”(claims 2,12), “generating an externally augmented feature, the externally augmented feature comprising one or more of a weather feature and an event feature.” (claims 3,13), “inputting the feature vector into a predictive model and inputting the feature vector into a neural network.” (claims 4,14,19), “inputting the feature vector into a decision tree-based model.” (claim 5, 15, 20), “inputting the feature vector into a Light GBM model.” (claims 6), “inputting the output of the neural network into a self-attention network and using an output of the self-attention network as a prediction in the plurality of predictions.” (Claims 7 inputting the feature vector into a recurrent neural network.” (claims 8,16, 20), “inputting the feature vector into a long-short term memory network.” (claims 9), and inputting the feature vector into a long-short term memory network.” (claims 10 and 17). These additional limitations (in claims 2-10, 12-17, and 18-20) also constitute concepts performed in the human mind which fall within the “Mental Processes” groupings of abstract ideas.
This judicial exception is not integrated into a practical application. Additional elements “computer readable medium comprising: computer program code (in claims 2-10, 12-17, and 18-20), all amount to no more than adding insignificant extra-solution activity/specifications related to data gathering, data input, or data transmittal. These additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The dependent claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of non-transitory computer readable medium comprising: computer program code are again insignificant extra-solution activity steps that cannot provide an inventive concept. All of these additional elements as generically claimed are considered well-understood, routine, and conventional.
Therefore, these limitations, taken alone or in combination, do not integrate the abstract idea into a practical application or recite significantly more that the abstract idea. Thus, all of the dependent claims are also not patent eligible.
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.
Claims 1-2, 4-12 and 14-20 are rejected under 35 U.S.C. 103 as being unpatentable over Zhuang (US 20210224648 A1: Pub. Date: 2021-07-22) in view of BOUILLET (Pub. No.: US
US 20180089582 Al, Pub. Date: 2018-03-29.) in further view of Saripalli ( US 20200337648 A1, 2020-10-29)
Regarding independent Claim 1,
Zhuang teaches a method comprising:
generating a feature vector by applying a feature engineering pipeline to raw data from a data warehouse (see Zhuang: Fig.6, [0103], “feature vectors may be generated”, Fig.5, [0095], “modeling server 504 may generate a plurality of feature vectors for each day.”), […]
the feature vector comprising a set of raw features including context-specific data associated with a time period and a location identifier (see Zhuang: Fig.6, [0102], “at step 602, event data in a sample time period may be received. For example, the modeling server 504 may receive event data associated with a plurality of events (e.g., transactions) occurring in a sample time period.” … [0094], “data associated with an identifier of a country associated with the PAN, data associated with a response code, data associated with a merchant identifier (e.g., a merchant name, a merchant location, and/or the like), data associated with a type of currency corresponding to funds stored in association with the PAN, and/or the like.”), and
a plurality of lag features generated using a sliding-window approach over past time periods (see Zhuang: Fig.2, [0075], “the model may sample event data from a sample time period (e.g., a week) based on event data of a series of subperiods (e.g., hours) that may be grouped (e.g., into days). The depicted example of the model illustrates the aggregation of event data over a series of subperiods (e.g., 168 hours, T.sub.0˜T.sub.167), and the prediction of values of a next series of subperiods (e.g., 24 hours, T.sub.168˜T.sub.191). In step 202, the model receives and aggregates event data (e.g., in a feature vector format, such as having aggregate values V.sub.1, V.sub.2, V.sub.3, V.sub.4, etc.) for each subperiod (e.g., each hour T.sub.n).”)
the plurality of lag features including a current value of a selected feature in the set of raw features and one or more historical values of the selected feature (see Zhuang: Fig.2, [0073], “Event data in real-world applications may be composed of multiple features. In the example of transaction data, features may include, but not limited to, transaction amount, number of transactions in the past hour, and/or the like. For a given period of time (e.g., a week), each feature may be represented as a time series. FIG. 1 depicts an example of event data evaluated as a time series, broken down by seven days over the course of a week, illustrating the patterns of change in features of the event data. Historic feature values of event data is represented by the “Past” section of the illustration, and such historic event data is used as an input to the described models herein.”)
inputting the feature vector into a plurality of base models of different model types (see Zhuang: Fig.6, [0104], “feature vectors may be provided as inputs to a set of first RNN models (base models). For example, the modeling server 504 may provide the plurality of feature vectors as inputs to a set of first recurrent neural network (RNN) models. The set of first RNN models may include a first plurality of nodes.”),
the plurality of base models including at least a decision tree-based model comprising a gradient-boosting model (see Zhuang: Fig,2, [0131], “Random forest is an ensemble method utilizing both bootstrap aggregating and random subspace methods for training a set of decision trees. The input to the model was once again a 679-sized vector consisting of the four time series from the last 168 hours, and the output of the model was a 96-sized vector consisting of the four time series for the next 24 hours. The hyper-parameters associated with random forest were found based on the estimated error of three-fold cross-validation.”) and a recurrent neural network model comprising a long short-term memory network having an LSTM portion that receives the plurality of lag features and generates a first prediction (see Zhuang: Fig,2, [0132], “RNN is an artificial neural network model for modeling sequence data. We used one or two layers of RNN to encode the input time series, then we used a Multi-layer Perceptron (MLP) to predict the time series for the next 24 hours. We tested the model with both long short-term memory (LSTM) and gated recurrent unit (GRU) recurrent architecture.”),
wherein the first prediction is concatenated with the feature vector and input into one or more dense layers to generate a neural network prediction (see Zhuang: Fig,2, [0076], “In step 206, the model merges the layers of each RNN 204A-204N. Each merge layer is configured to aggregate the event data from each subperiod (e.g., hour) in the sample time period (e.g., week). In step 208, the model inputs the merged layers of step 206 into an additional RNN. The additional RNN connects to the previous merge layers to capture the high-level patterns in the sample time period. In step 210, the model outputs the fully connected layers to make predictions for each feature of the event data individually.”)and outputting a plurality of predictions (see Zhuang: Fig.6, [0105], “At step 608, first outputs from the set of first RNN models may be generated.”),
each prediction in the plurality of predictions representing future values of the selected feature, each base model processing the same feature vector to generate its respective prediction (see Zhuang: Fig,2, [0080], “The model utilizes h.sub.s.sup.(t) to predict the future values of each future subperiod (e.g., hour) by augmenting fully connected layers to RNN.sub.s, in step 210. Instead of predicting values of all features at once, the model may utilize separated linear fully connected layers to predict every feature individually. These fully connected layers may be represented by the following formula: v.sub.k.sup.(t)=W.sub.ks.sub.t+b.sub.k Formula 4, where W.sub.k is the weight matrix and b.sub.k is the bias term of the fully connected layer.”)
predicting a future value of the selected feature by inputting the plurality of predictions into a meta-model (see Zhuang: Fig.6, [0108], “final outputs from the second RNN model may be generated. For example, the modeling server 504 may generate final outputs for each RNN node of the second RNN model, such as by generating a final hidden layer feature vector for a hidden layer of each RNN node of the second RNN model based on one of the aggregated time-series layers that was provided as an input to a respective RNN node. Each of the final outputs may include a final hidden layer feature vector.”). See also stating [0080], “model utilizes h.sub.s.sup.(t) to predict the future values of each future subperiod (e.g., hour) by augmenting fully connected layers to RNN.sub.s, in step 210. Instead of predicting values of all features at once, the model may utilize separated linear fully connected layers to predict every feature individually. These fully connected layers may be represented by the following formula:”), […]
controlling a computing system based on the predicted future value of the selected feature, wherein the computing system includes at least one of a fraud detection system, an inventory management system, or a resource allocation system (see Zhuang: Fig.8, [0116], “at least one parameter of the fraud detection system 509 may be altered. For example, the fraud detection system 509 may alter one or more parameters based on the final outputs of the second RNN model. Parameters of the fraud detection system 509 may include, but are not limited to, transaction account or merchant transaction volume thresholds for executing fraud prevention processes, authorized transaction accounts, unauthorized transaction accounts, authorized merchants, unauthorized merchants, transaction value limit thresholds for executing fraud prevention processes, transaction count thresholds for merchants or transaction accounts for executing fraud prevention processes, parameters of an automatic network communication system 506 for transmitting fraud alerts, and/or the like.”), […]
Zhuang does not teach the system wherein:
the meta-model comprising a linear regression model that applies a learned weighting to each of the plurality of predictions, the meta-model applying a learned weighting to each of the plurality of predictions to generate an ensemble prediction with improved accuracy compared to any individual base model prediction by combining predictions from the different model types to reduce prediction variance and improve generalization performance across different data distributions; and
controlling comprises automatically adjusting operational parameters of the computing system to respond to the predicted future value.
the feature engineering pipeline comprising cleaning the raw data by removing outliers and interpolating missing data to produce cleaned data and augmenting the cleaned data.
However, BOUILLET teaches the system wherein:
meta-model comprising a linear regression model that applies a learned weighting to each of the plurality of predictions (see BOUILLET: Fig.1, [0017], “A model may be a specific instance of a model class (e.g., a regression model for category “A”). A model version may be a specific version of a model. An ensemble model may be a model that combines predictions of one or more bases models. A policy may be one or more parameters that may define an ensemble model, such as, for example, what models (e.g., model types or classes) are included in the ensemble model and one or more associated parameters, weights that may be associated with each base model prediction, and which models may be rerun, retained, rebuilt.”)
the meta-model applying a learned weighting to each of the plurality of predictions to generate an ensemble prediction with improved accuracy compared to any individual base model prediction (see BOUILLET: Fig.5, [0058], “A template model 502 is shown containing, in the depicted embodiment, model classes 1-4 that are associated with an ensemble model (i.e. meta-model) having at least one policy. In conjunction with the deployment model 504 is a depiction of the template model deployed as a deployment model 504. The model classes 1-4 input respective forecast predictions into a family of ensemble models, as indicated in deployment model 504, while also feeding a policy manger 506 with predicted target variables, sampled ground truth data, model definitions, and/or contextual information” … [0059], “The policy manager 506, in communication with each one of the ensemble modules, may include logic to analyze each ingested streaming data received into the ensemble models from the model classes and also analyze the predicated target variables of the ensemble models. The policy manger 506 may identify one or more error states according to each prediction stream of the ensemble models.”), by combining predictions from the different model types to reduce prediction variance and improve generalization performance across different data distributions (see BOUILLET: Fig.5, [0061], “the mechanisms of the embodiment enable dynamic (e.g., real-time) adjustment of the family of ensemble models, provide increased predictive accuracy for target variables, and leverage the most current and available error state information on constituent models to improve the prediction stream of one or more ensemble models based upon a prediction stream of another ensemble model. Error states may be correlated across ensemble models that have substantial similarity (e.g., more likely than not similar or have more similar parameters than non-similar parameters) so to increase efficiency for managing a large family of models with a large number of tunable parameters.”); and
controlling comprises automatically adjusting operational parameters of the computing system to respond to the predicted future value (see BOUILLET: Fig.5, [0059], “policy manager 506 may derive a plurality of policies for the family of ensemble models to predict a plurality of target variables for the incoming, streaming data such that the plurality of policies enables dynamic adjustment of the prediction system. The policies may be generated according to contextual information, model similarities, model definitions, prediction quality metrics, ground truth data, or a combination thereof. The policy manager 506 may dynamically update one or more of the policies according to one or more error states of the set of ensemble models.”)
Because both Zhuang and BOUILLET are in the same/similar field of endeavor of copy and paste operation and clipboard management, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention, to modify the system of Zhuang to include the system, wherein model applying a learned weighting to each of the plurality of predictions to generate an ensemble prediction with improved accuracy compared to any individual base model prediction and automatically adjusting operational parameters of the computing system to respond to the predicted future value as taught by BOUILLET. One would have been motivated to make such a combination in order to improve the prediction stream of one model based upon a prediction stream of another model in the family of model. (see BOUILLET [0017]).
Zhuang and BOUILLET does not teach the method wherein the feature engineering pipeline comprising cleaning the raw data by removing outliers and interpolating missing data to produce cleaned data and augmenting the cleaned data.
However, Saripalli teaches the method wherein:
the feature engineering pipeline comprising cleaning the raw data by removing outliers and interpolating missing data to produce cleaned data and augmenting the cleaned data (see Saripalli: Fig.12, [0155], “The processing method can be a bottom-up processing method or a top-down processing method, for example. When the processing method is to be a bottom-up processing method, at block 1206, the data is cleaned. For example, the data can be cleaned by the data processor 1130 to normalize the data with respect to other data and/or a reference/standard value. The data can be cleaned by the data processor 1130 to interpolate missing data in the time series, for example. The data can be cleaned by the data processor 1130 to adjust a format of the data, for example. At block 1208, outliers in the data are identified and filtered. For example, outlier data points that fall beyond a boundary, threshold, standard deviation, etc., are filtered (e.g., removed, separated, reduced, etc.) from the data being processed”).
the feature vector comprising a set of raw features including context-specific data associated with a time period and a location identifier and a plurality of lag features generated using a sliding-window approach over past time periods (see Saripalli: Fig.12, [0155],
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention, to modify the system of Zhuang to include the system, wherein the feature engineering pipeline comprising cleaning the raw data by removing outliers and interpolating missing data to produce cleaned data and augmenting the cleaned data as taught by Saripalli. One would have been motivated to make such a combination in order to improve the prediction by providing quality and complete data set to the machine learning model.
Regarding Claim 2,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 1. Zhuang further teaches generating the feature vector (see Zhuang: Fig.6, [0103], “At step 604, feature vectors may be generated”), further comprises:
generating an intrinsically augmented feature, the intrinsically augmented feature comprising one or more of a synthetic date feature, a historical feature, and an aggregate feature (see Zhuang: Fig.1, [0073], “FIG. 1 depicts an example of event data evaluated as a time series, broken down by seven days over the course of a week, illustrating the patterns of change in features of the event data. Historic feature values of event data is represented by the “Past” section of the illustration, and such historic event (a historical feature) data is used as an input to the described models herein. Future feature values of event data is represented by the “Future” section of the illustration, and such future event data (aggregate feature) is generated as an output from the described models herein.”)
Regarding Claim 4,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 1. Zhuang further teaches the method wherein inputting the feature vector into the plurality of base models (see Zhuang: Fig.6, [0104], “feature vectors may be provided as inputs to a set of first RNN models), further comprises:
inputting the feature vector into a predictive model and inputting the feature vector into a neural network (see Zhuang: Fig.1, [0074], “multi-stream recurrent neural network (RNN) model to provide for future event prediction, e.g., merchant transaction prediction. As described herein, the method includes aggregating event data at a fine granularity (e.g., day) and then augmenting the event data with another RNN to capture the pattern at a higher level of granularity. Described systems and methods summarize and preserve as much event information as possible for future long-term predictions.”)
Regarding Claim 5,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 4. Zhuang further teaches the method wherein:
inputting the feature vector into the predictive model comprises inputting the feature vector into a decision tree-based model (see Zhuang: [0131], “utilizing both bootstrap aggregating and random subspace methods for training a set of decision trees. The input to the model was once again a 679-sized vector consisting of the four time series from the last 168 hours, and the output of the model was a 96-sized vector consisting of the four time series for the next 24 hours. The hyper-parameters associated with random forest were found based on the estimated error of three-fold cross-validation.”)
Regarding Claim 6,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 5. Zhuang further teaches the method wherein:
inputting the feature vector into the decision tree-based model comprises inputting the feature vector into a Light-GBM model (see Zhuang: [0131], “Random Forest is an ensemble method utilizing both bootstrap aggregating and random subspace methods for training a set of decision trees. The input to the model was once again a 679-sized vector consisting of the four time series from the last 168 hours, and the output of the model was a 96-sized vector consisting of the four time series for the next 24 hours.”. Examiner notes that both Random Forest and Light-GBM are tree-based ensemble methods.)
Regarding Claim 7,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 4. Zhuang further teaches the method wherein:
inputting the feature vector into the plurality of base models further comprising inputting the output of the neural network into a self-attention network (see Zhuang: Fig.2, [0076], “a first RNN 204A is applied for the event data of the first day (e.g., T.sub.0˜T.sub.23), a second RNN 204B is applied for the event data of a second day (e.g., T.sub.24˜T.sub.47), and an nth RNN 204N is applied for the event data of an nth day (e.g., T.sub.144˜T.sub.167.”), and using an output of the self-attention network as a prediction in the plurality of predictions [0076], “the model merges the layers of each RNN 204A-204N. Each merge layer is configured to aggregate the event data from each subperiod (e.g., hour) in the sample time period (e.g., week). In step 208, the model inputs the merged layers of step 206 into an additional RNN. The additional RNN connects to the previous merge layers to capture the high-level patterns in the sample time period. In step 210, the model outputs the fully connected layers to make predictions for each feature of the event data individually.”)
Regarding Claim 8,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 4. Zhuang further teaches the method wherein:
inputting the feature vector into the neural network comprises inputting the feature vector into a recurrent neural network (see Zhuang: Fig.5, [0096], “The modeling server 504 may provide the plurality of feature vectors as inputs to a set of first recurrent RNN models.”)
Regarding Claim 9,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 7. Zhuang further teaches the method wherein:
inputting the feature vector into the recurrent neural network comprises inputting the feature vector into a long-short term memory network (see Zhuang: Fig,2, [0132], “RNN is an artificial neural network model for modeling sequence data. We used one or two layers of RNN to encode the input time series, then we used a Multi-layer Perceptron (MLP) to predict the time series for the next 24 hours. We tested the model with both long short-term memory (LSTM) and gated recurrent unit (GRU) recurrent architecture.”)
Regarding Claim 10,
As shown above, Zhuang, BOUILLET and Saripalli teaches all the limitations of Claim 8. Zhuang further teaches the method wherein inputting the feature vector into the long- short term memory network comprises:
generating a first prediction of the selected feature by inserting the plurality of lag features into the long-short term memory network (see Zhuang: Fig.2, [0075], “The depicted example of the model illustrates the aggregation of event data over a series of subperiods (e.g., 168 hours, T.sub.0˜T.sub.167), and the prediction of values of a next series of subperiods (e.g., 24 hours, T.sub.168˜T.sub.191). In step 202, the model receives and aggregates event data (e.g., in a feature vector format, such as having aggregate values V.sub.1, V.sub.2, V.sub.3, V.sub.4, etc.) for each subperiod (e.g., each hour T.sub.n).”;
combining the first prediction with the feature vector to generate a concatenated vector (see Zhuang: Fig.2, [0076], “the model inputs the merged layers of step 206 into an additional RNN. The additional RNN connects to the previous merge layers to capture the high-level patterns in the sample time period. In step 210, the model outputs the fully connected layers to make predictions for each feature of the event data individually.”). See also [0129], stating “Linear model predicts a value by linearly combining the input vector. Since we predicted the four features for the next 24 hours (e.g., 96 values total), we trained 96 linear models and each model predicted one value. The input to each of the models was a 672-sized vector consisting of the four time series from the last 168 hours. We tested linear model under both L2-regularized and non-regularized settings. When L2-regularization was used, the parameter associated with the strength of regularization was found using three-fold cross-validation”); and
generating a second prediction of the selected feature by inserting the concatenated vector into one or more dense layers (see Zhuang: Fig.2, [0076], “The model outputs the fully connected layers to make predictions for each feature of the event data individually.”)
Regarding independent Claim 11,
Claim 11 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as Claim 1 and is rejected under the same rationale.
Regarding Claim 12,
Claim 12 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as claim 2 and is rejected under the same rationale.
Regarding Claim 14,
Claim 14 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as claim 4 and is rejected under the same rationale.
Regarding Claim 15,
Claim 15 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as Claim 5 and is rejected under the same rationale.
Regarding Claim 16,
Claim 16 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as Claim 8 and is rejected under the same rationale.
Regarding Claim 17,
Claim 17 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as Claim 10 and is rejected under the same rationale.
Regarding Claim independent 18,
Claim 18 is directed to a system claim and has the same/similar claim limitations as Claim 1 and is rejected under the same rationale.
Regarding Claim 19,
Claim 19 is directed to a system claim and has the same/similar claim limitations as Claim 4 and is rejected under the same rationale.
Regarding Claim 20,
Claim 20 is directed to a system claim and has the same/similar claim limitations as Claim 5 and Claim 8 and is rejected under the same rationale.
Claims 3 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Zhuang in view of BOUILLET and Saripalli as applied to Claims 1-2, 4-12 and 14-20 as shown above and in further view of Wang (Pub. No.: US 11367141B 1, Pat. Date: 2022-06-21)
Regarding Claim 3,
As shown above, Zhuang teaches all the limitations of Claim. Zhuang does not teach or disclose generating the feature vector further comprises generating an externally augmented feature, the externally augmented feature comprising one or more of a weather feature and an event feature.
However, Wang teaches the method wherein generating the feature vector further comprises generating an externally augmented feature, the externally augmented feature comprising one or more of a weather feature and an event feature (see Wang: Fig.11, Col.4, Line 49-61, “The classes of data used to augment internal data is referred to herein as “external data,” while together, the combination of internal and external datasets is herein referred to as the “360-Degree Database.” Examples of classes (or categories) of external data components of the 360-Degree Database include, but are not limited to, the following: Geo-Demographic, Geo-Economic, Geospatial, Traffic, Weather, Catastrophes and Natural Disasters, Public Health and Safety, Legal and Regulatory, Societal, Industry Benchmarks, Insured Asset Class data, Insured Party Financial data, and Insured Party Behavioral data. The term “insured party” can represent an individual, a household, a business or other entity.”)
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Zhuang to include an externally augmented feature that comprise weather feature and an event feature as taught by Wang. One would have been motivated to make such a combination in order to provide accurate predictions by collecting and assembling relevant data to generate a forecasting model.
Regarding Claim 13,
Claim 13 is directed to a non-transitory computer-readable storage medium and has the same/similar claim limitations as claim 3 and is rejected under the same rationale.
Response to Arguments
Claim Rejections - 35 U.S.C. § 101,
Regarding the 35 U.S.C. 101 rejection for being directed non-statutory subject matter has been updated based on applicant amendments and. Therefore, the 35 U.S.C. 101 rejection has been sustained.
Claim Rejections - 35 U.S.C. § 103,
Applicant’s arguments with respect to claim amendments have been considered but are moot considering the new combination of references being used in the current rejection. The new combination of references was necessitated by Applicant’s claim amendments. Therefore, the claims are rejected under the new combination of references as indicated above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
PGPUB
NUMBER:
INVENTOR-INFORMATION:
TITLE / DESCRIPTION
US 20230205740 A1
ABDELAAL, MOHAMED
Title: META-LEARNING SYSTEMS AND/OR METHODS FOR ERROR DETECTION IN STRUCTURED DATA
Description: Techniques and systems described herein improve the technical field of machine learning by providing more flexibility in model training, as compared to current training methods. For example, the techniques and systems described herein allow for “transforming” a machine learning model from one type to another type by training a particular type of machine learning model to mimic another type of machine learning mode.
US 20230281518 A1
Verma; Dinesh C.
Title: DATA SUBSET SELECTION FOR FEDERATED LEARNING
Description: The present application relates generally to a multi-part meta-learning approach to detecting errors in potentially large and complex datasets including structured data..
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ZELALEM W SHALU whose telephone number is (571)272-3003. The examiner can normally be reached M- F 0800am- 0500pm.
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, Cesar Paula can be reached on (571) 272-4128. 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.
/Zelalem Shalu/Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights