Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
This Non-Final Rejection is filed in response to Request for Continued Examination filed 05/08/2026
Claims 1, 19, and 20 are amended.
Claims 1-11 and 13-20 remain pending.
Response to Arguments
Argument 1, Applicant argues in Applicant Arguments/Remarks Made in an Amendment filed 05/08/2026, pg. 12, that prior art fails to teach the primary claim limitations, “receiving, the first model, at least one modification to at least one default artificial intelligence model parameter value inherited by the first model from the scenario settings;”
Response to Argument 1, the examiner respectfully disagrees and upon further examination of the art along with remapping of the primary claim limitations, the examiner notes that Sarferaz teaches that a user may customize a default template algorithm within a predictive analytic library (PAL) as part of adopting the template for their specialized field of use. The examiner notes that the BRI for, “receiving, the first model, at least one modification to at least one default artificial intelligence model parameter value inherited by the first model from the scenario settings”, would encompass the modifications to a base template algorithm selected by the user as part of a predictive sales scenario. This is supported by para. [0052] of Sarferaz:
[0052], For development efficiency for each PAL algorithm in the PAL library 508, a corresponding template ABAP AMDP class 502 is implemented which has a pre-defined interface consisting of, for example, a train method and a consume method. In one example, a template is provided for each algorithm provided in the PAL library 508 (e.g., decision tree, ransom forest). For extensibility, business add-ins (BADI) enhancement points are provided for those methods. An example template 502 is shown in FIG. 6 as an ABAP class template 600 for a data structure definition, FIG. 7 is a template 700 including a training method 702 and a consumption method 704, and FIG. 8 is a template 800 for implementation.
Argument 2, Applicant argues in Applicant Arguments/Remarks Made in an Amendment filed 05/08/2026, pg. 12-13, that prior art fails to teach the primary claim limitations, “combining in first model parameters for the first model, the at least one modification to the at least one default artificial intelligence model parameter value inherited by the first model from the scenario settings and other artificial intelligence model parameter values inherited by the first model from the scenario setting”
Response to Argument 2, the examiner respectfully disagrees and upon further examination of the art along with remapping of the primary claim limitations, the examiner notes that Sarferaz teaches that the modifications made by the developer are used in combination with the previously selected sales prediction scenario settings. Wherein the BRI for the limitation, “combining in first model parameters for the first model, the at least one modification to the at least one default artificial intelligence model parameter value inherited by the first model from the scenario settings and other artificial intelligence model parameter values inherited by the first model from the scenario setting”, encompasses how the template is combined with user updated customization values as part of the user adopting the template for their specific scenario. This is supported by para. [0052] of Sarferaz:
[0052] The templates shown in FIGS. 6-8 are templates for the PAL algorithms random decision trees (RDS). Application developers can copy and adopt those templates for the specific ML scenario to code the logic for training and the logic for consumption. For example, following objects can be copied and adopted (e.g., see coding sections in the templates in FIGS. 6-8 marked with “<<<<<<TODO:”): [0053] 1. Class CL_HEMI_TEMPLATE_RDT [0054] 2. Interface for BAdi: IF_HEMI_TEMPLATE_BADI_RDT [0055] 3. Enhancement Spot ES_HEMI_TEMPLATE_RDT [0056] 4. Fallback Classes CL_HEMI_TEMPLATE_FLBK_RDT and CL_HEMI_TEMPLATE_FLBK2_RDT
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) 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.
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.
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.
Claim(s) 1-5, 11, 14, 17, & 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz” in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”.
Claim 1:
Sarferaz teaches a computer-implemented method comprising: receiving scenario settings (i.e. para. [0060], “A user (e.g., a data scientist) associated with the entity can determine which fields (e.g., of sales information data) are relevant for such a use case… The user enters this information (e.g., via an application or user interface 202, 302, 304 or modelling and administration 310 interface) and the computing system receives the information”, wherein the BRI for scenario settings encompasses data associated with predictive scenario, such as a sales use case scenario setting) for a predictive scenario (i.e. para. [0058], “PAI predictive scenario 902 is also created for the PAL based scenarios. Predictive scenario 902 is a business application specific ML design time artifact which includes metadata regarding application data source (e.g. table or view) for training, algorithm type (e.g. regression or classification) and consumption API”, wherein the BRI for scenario settings encompasses the predictive scenario artefacts) for a target field of a dataset (i.e. para. [0058], “Predictive scenario 902 is a business application specific ML design time artifact which includes metadata regarding application data source (e.g. table or view) for training,”, wherein a sales setting for a predictive scenario setting may be for specific data sources associated with predicting future sales algorithm models stored in the predictive analytics library (PAL) and automated predictive library (APL)) , wherein the scenario settings include automatically determined default artificial intelligence model parameter values (i.e. para. [0058], “Predictive scenario 902 is a business application specific ML design time artifact which includes metadata regarding application data source (e.g. table or view) for training, algorithm type (e.g. regression or classification) and consumption API. ”, wherein the BRI for automatically determined default AI model parameter values encompasses the sales data set definition, the target fields, and the potential algorithms API together within the predictive scenario) are applicable to multiple artificial intelligence models that are able to implement the predictive scenario (i.e. para. [0014], “use cases such as forecasting, key influencers, and trending can be solved with classic machine learning algorithms like regression, clustering, classification, or time series analysis”, wherein a predictive sales scenario is applicable to multiple predictive models, such as regression, clustering, classification, or time series analysis);
Presenting a list of the multiple artificial intelligence models (i.e. para. [0041], “The intention is to harmonize the management of ML models 312 and to provide a simple common interface to allow applications to interact with different types of supported ML libraries without the requirement for applications to develop ML-engine-specific code”, wherein the BRI for a list of models encompasses a display of a library repository of relevant algorithms) that are able to implement the predictive scenario (i.e. para. [0015], “the client device 110 may comprise a display module (not shown) to display information (e.g., in the form of user interfaces)”, wherein it is noted that the algorithms may be Predictive Analytics);
Receiving a selection, from the list of the multiple artificial intelligence models that are able to implement the predictive scenario, of a first model (i.e. para. [0044], The ML model 312 can then be consumed via the CDS view for ML 318 which accesses the ML model 312 via an API (e.g., the PAI: repository, engine adapter, and API 326). The API inputs data required for the ML model 312, receives the output of the ML model 312, and returns the output to the CDS view for ML 318) ;
in response to receiving the selection of the first model, inheriting, by the first model and from the scenario settings (i.e. para. [0050], “ if one of the attributes (e.g., revenue prediction) corresponds to a ML model 312, the ABAP class 406 calls the SQL script procedure 402 to access the machine learning model 312 to calculate and output the data for the attribute (e.g., revenue prediction)”, wherein the once the model template is selected the core data service (CDS) will identify and format the specific sales data prediction attributes from the sales scenario as part of formatting the default template for predictive sales scenario), the default artificial intelligence model parameter values included in the scenario settings (i.e. para. [0052], “For development efficiency for each PAL algorithm in the PAL library 508, a corresponding template ABAP AMDP class 502 is implemented which has a pre-defined interface consisting of, for example, a train method and a consume method. In one example, a template is provided for each algorithm provided in the PAL library 508 (e.g., decision tree, ransom forest). “, wherein selection of a model results in inheriting the default template settings and data scenario for the model);
receiving, the first model, at least one modification to at least one default artificial intelligence model parameter value inherited by the first model from the scenario settings (i.e. para. [0052], “a corresponding template ABAP AMDP class 502 is implemented which has a pre-defined interface consisting of, for example, a train method and a consume method. In one example, a template is provided for each algorithm provided in the PAL library 508 (e.g., decision tree, ransom forest). )”, wherein the BRI for a model setting encompasses an attribute used to train the selected model);
combining in first model parameters for the first model, the at least one modification to the at least one default artificial intelligence model parameter value inherited by the first model from the scenario settings and other artificial intelligence model parameter values inherited by the first model from the scenario setting (i.e. para. [0052], “As an example, the templates shown in FIGS. 6-8 are templates for the PAL algorithms random decision trees (RDS). Application developers can copy and adopt those templates for the specific ML scenario to code the logic for training and the logic for consumption.”, wherein the template is combined with user updated values as part of the user adopting the template for their specific scenario); processing a copy of the dataset based on the first model parameters to generate a prepared dataset (i.e. para. [0061], ”the computing system accesses application data to generate data corresponding to the specified fields to train the first machine learning model using the specified machine learning algorithm. For example, the computing system accesses application data or tables 324 and puts this data to the machine learning algorithm to train the first machine learning model”, wherein the BRI for a copy encompasses accessing data from a specified dataset and inputting a copy of the data into the customized machine learning algorithm); providing the prepared dataset and the first model parameters to a predictive analytical library that is configured to build, train, and test artificial intelligence models (i.e. para. [0066] “the SQL script procedure 402 and corresponding CDS table function 404 and ABAP class 406 can be generated and used for breakout scenarios such as machine learning models based on machine learning algorithms in the PAL library”, wherein library of machine learning algorithms use application data as input for machine learning model building, training, and optimization); receiving, from the predictive analytics library, a reference to a first trained artificial intelligence model trained by the predictive analytical library based on the prepared dataset and the first model parameters (i.e. para. [0058], As the PAI training app 904 is based on predictive scenario 902, it triggers the training method of the AMDP class 906 and saves the trained model 312. ML applications 910 can consume the trained model 312 via the AMDP class 906 or via API of the predictive scenario 902) q (i.e. para. [0075], “the computing system executes the CDS view for ML 318 to generate data for input to the first machine learning mode”, wherein a trained first machine learning model may be called on and executed for its intended predictive analysis scenario); receiving a request to generate a prediction for the predictive scenario for the target field for at least one record of the dataset (i.e. para. [0075], the CDS view for ML 318 defines what data is needed to input to the first machine learning model, and accesses one or more data stores (e.g., application data 210, application data or tables 324) to generate the data. The computing system inputs the generated data into the first machine learning model); providing the at least one record of the dataset to the first trained artificial intelligence model (i.e. para. [0075], accesses one or more data stores (e.g., application data 210, application data or tables 324)); receiving, from the first trained artificial intelligence model, a prediction for the target field for each record of the at least one record of the dataset (i.e. para. [0075], an API (e.g., PAI: repository, engine adapter, and API 326) and receives the output (e.g., prediction) from the first machine learning model); and providing at least one prediction for presentation in a user interface that displays information from the dataset (i.e. para. [0044], The API inputs data required for the ML model 312, receives the output of the ML model 312, and returns the output to the CDS view for ML 318. The output can be made available via an application or user interface 302).
While Sarferaz teaches using a predictive analytics library to store trained models based on a user prepared dataset that has first model parameters, Sarferaz may not explicitly teach
first model evaluation data that reflects model performance of the first model for predicting the target field of the dataset;
Receiving a selection, from the list of the multiple artificial intelligence models that are able to implement the predictive scenario, of a second model for comparison to the first model, wherein the second model is:
trained using model settings for the second model that are different from the model settings of the first model; and
used to generate another prediction for the predictive scenario that is different from predictions received from the first model.
However, Schmidt teaches
first model evaluation data that reflects model performance of the first model for predicting the target field of the dataset (i.e. para. [0035], “At some point, the threshold value suggested by the predictive model may be more accurate than the value supplied by the heuristic. Accordingly, when a user determines that the predictive model has been trained to an appropriate degree, the user can change decision 154 to operate using rules based on predictive analysis”, wherein the BRI for evaluation data encompasses a predictive accuracy metric that reflects the degree to which a model performs in predicting the database field associated with the predictive model).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add first model evaluation data that reflects model performance of the first model for predicting the target field of the dataset, to Sarferaz’s user construction and training of specified predictive models, with how a predictive model also has an accuracy evaluation metric that reflects the accuracy performance of a predictive model, as taught by Schmidt. One would have been motivated to combine Schmidt with Sarferaz, and would have had a reasonable expectation of success, as the combination provides a user with more information regarding evaluations that need to be taken in order to advance the process.
While Sarferaz-Schmidt-Matlab teach a library of models for user selection and modification and implies that the PAI repository library includes support for model comparison to allow assessing metrics, Sarferaz-Schmidt may not explicitly teach
Receiving a selection, from the list of the multiple artificial intelligence models that are able to implement the predictive scenario, of a second model for comparison to the first model, wherein the second model is:
trained using model settings for the second model that are different from the model settings of the first model; and
used to generate another prediction for the predictive scenario that is different from predictions received from the first model.
However, Matlab teaches
Receiving a selection, from the list of the multiple artificial intelligence models that are able to implement the predictive scenario, of a second model for comparison to the first model (i.e. Wherein it is noted in Figs. 1-5, that multiple models may be used to implement the predictive scenario and may be compared against each other in the history window of the data browser), wherein the second model is:
trained using model settings for the second model that are different from the model settings of the first model (i.e. Wherein it is noted in Figs. 6-7, that a user may change the data set prediction variables before selecting a different algorithm model); and
used to generate another prediction for the predictive scenario that is different from predictions received from the first model (i.e. Wherein it is noted in Figs. 1-7 that a different accuracy prediction level for the data is generated that may be different from a first model due to a separate algorithmic model being used on the same data).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add Receiving a selection, from the list of the multiple artificial intelligence models that are able to implement the predictive scenario, of a second model for comparison to the first model, wherein the second model is: trained using model settings for the second model that are different from the model settings of the first model; and used to generate another prediction for the predictive scenario that is different from predictions received from the first model, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with the user interface for editing and comparing constructed machine learning models, as taught by Matlab. One would have been motivated to combine Matlab with Sarferaz-Schmid, and would have had a reasonable expectation of success, as the combination results in a user being able to better calculate optimal computational efficiency and best fit for specific business requirements.
Claim 2:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1, wherein the first trained artificial intelligence model is a trained machine learning model (i.e. para. [0036], a modelling and administration interface 310 can be used to access the third-party server system 130 via SAP S/4HANA to train machine learning models and generate trained machine learning models (e.g., ML model 312)).
Claim 3:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz further teaches wherein the scenario settings include a specification of the dataset and the target field (i.e. para. [0060], user (e.g., a data scientist) associated with the entity can determine which fields (e.g., of sales information data) are relevant for such a use case).
Claim 4:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz further teaches wherein the scenario settings include at least one filter condition for the target field (i.e. para. [0060], “an entity may wish to provide functionality to its users to forecast revenue for a specified product, as one example use case. In order to provide this functionality, the entity can use a machine learning model to learn from existing sales order data and predict future revenue for a specified product”, wherein the BRI for a filter condition encompasses filtering out data not related to a targeted and specific product field).
Claim 5:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz further teaches wherein the scenario settings include indications of fields of the dataset to include in training of artificial intelligence models of the predictive scenario (i.e. para. [0060], “The user enters this information (e.g., via an application or user interface 202, 302, 304 or modelling and administration 310 interface) and the computing system receives the information”, wherein the BRI for indications encompasses displayed user selected information on an API).
Claim 11:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Matlab further teaches wherein each model of the at multiple intelligence model is trained to predict values of the target field (i.e. wherein it is noted in Fig. 7 that each model of the multiple selected models are trained to predict an accuracy value).
Claim 14:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz further teaches wherein a first model-specific setting for the first model is different from a second model-specific setting for a second model of the predictive scenario (i.e. para. [0038], “a virtual data model (VDM) is implemented using CDS views 316. One purpose of these VDM views is to hide the cryptic database model and to provide a reusable semantic layer which can be consumed in different scenarios, e.g. analytics, planning, search, or transactions”, wherein it is noted that different settings may be used for different scenarios).
Claim 17:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Schmidt further teaches wherein each prediction of the target field includes a predicted outcome of the target field and a prediction probability for the predicted outcome (i.e. para. [0039], FIG. 1 can illustrate multiple execution steps suggested by the predictive model, optionally with an indication of the relative confidence or probability of the potential steps).
Claim 19:
Claim 19 is the system claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Sarferaz further teaches
one or more computers (i.e. para. [0015], Fig. 1, he client device 110 may comprise, but is not limited to, a mobile phone, desktop computer); and
a computer-readable medium coupled to the one or more computers having instructions stored thereon which, when executed by the one or more computers, cause the one or more computers to perform operations (i.e. para. [0099], the machine-readable medium 1338 is non-transitory (in other words, not having any transitory signals) in that it does not embody a propagating signal).
Claim 20:
Claim 20 is the product claim reciting similar limitations to claim 1 and is rejected for similar reasons.
Claim(s) 6-7 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz” in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, as applied to Claim 1 above, and further in light of U.S. Patent Application Publication NO. 20160267396 “Gray”.
Claim 6:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz, Schmidt, and Matlab may not explicitly teach
wherein the scenario settings include default data pre-processing settings for pre-processing the dataset before artificial intelligence models of the predictive scenario are trained.
However, Gray teaches
wherein the scenario settings include default data pre-processing settings for pre-processing the dataset before artificial intelligence models of the predictive scenario are trained (i.e. para. [0035]. “the workflow auditing system 136 includes one or more data sources associated with a complex processing workflow that allow input of different types of data or information (automated and non-automated) related to a complex processing task to be provided or input to the training server 102 and/or the prediction/scoring server 108”, wherein the BRI for settings encompasses the automated pre-processing of data before it is provided to a prediction server). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the scenario settings include default data pre-processing settings for pre-processing the dataset before artificial intelligence models of the predictive scenario are trained, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with how the default setting for pre-processing data is to automatically perform some sort of processing on data before it is provided to a server, as taught by Gray. One would have been motivated to combine Gray with Sarferaz-Schmidt-Matlab, and would have had a reasonable expectation of success, as the combination saves a user time and helps create a more accurate model.
Claim 7:
Sarferaz, Schmidt, Matlab, and Gray teach the computer-implemented method of claim 6.
Gray further teaches wherein the default data pre-processing settings includes settings for removal of outlier values and null values from the dataset (i.e. para. [0068], data preprocessing may include data cleaning, removal of outliers, identifying and treating missing values, and transformation of values, etc).
Claim(s) 8-9 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz” in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, as applied to Claim 1 above, and further in light of U.S. Patent Application Publication NO. 20150227520 “Clark”.
Claim 8:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz, Schmidt, and Matlab may not explicitly teach
wherein the scenario settings include default training settings for artificial intelligence models of the predictive scenario.
However, Clark teaches
wherein the scenario settings include default training settings for artificial intelligence models of the predictive scenario (i.e. para. [0024], The predictive algorithms stored in the predictive algorithms 110 may be generated by the QA application 112, or by a different source (such as specialized industry-standard prediction models).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the scenario settings include default training settings for artificial intelligence models of the scenario, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with how default training for a field is consistent with specialized industry standard for prediction models being applied to a certain type of predictive problem scenario, as taught by Clark. One would have been motivated to combine Gray with Sarferaz-Schmidt-Matlab, and would have had a reasonable expectation of success, as the combination saves a user time and assists in further improving the accuracy of any predictive models used to generate the candidate answers.
Claim 9:
Sarferaz, Schmidt, Matlab, and Clark teach the computer-implemented method of claim 8.
Clark further teaches
wherein the default training settings specify an artificial intelligence algorithm type and a specific artificial intelligence algorithm of the artificial intelligence algorithm type and wherein the default training settings specify default parameters for the specific artificial intelligence algorithm (i.e. para. [0024], “a predictive algorithm in the predictive algorithms 110 may be an industry-standard model that predicts consumer spending or the United States gross domestic product”, wherein the BRI for default encompasses the industry standard algorithms used in predictive models for predicting consumer spending).
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz”, in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, and further in light of U.S. Patent Application Publication NO. 20150227520 “Clark”, as applied to Claim 9 above, and further in light of U.S. Patent Application Publication NO. 20210241177 “Wang”.
Claim 10:
Sarferaz, Schmidt, Matlab, and Clark teach the computer-implemented method of claim 8.
Sarferaz, Schmidt, Matlab, and Clark may not explicitly teach
wherein the default training settings include a data split configuration that indicates a split between a training portion of the dataset and a test portion of the dataset.
However, Wang teaches
wherein the default training settings include a data split configuration that indicates a split between a training portion of the dataset and a test portion of the dataset (i.e. para. [0046-0047], “the automatic machine learning technology may relate to at least one of: an automatic data splitting for splitting the historical data into training data and verification data… Specifically, the historical data may be automatically split into the training data and the verification data according to a preset splitting rule”, wherein the BRI for default training settings include a data split configuration encompasses how, by default, data of a dataset is automatically split into at least a training a verification portion).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the scenario settings include default training settings for artificial intelligence models of the scenario, to Sarferaz-Schmidt-Matlab-Clark’s user construction and training of specified predictive models, with wherein the default training settings include a data split configuration that indicates a split between a training portion of the dataset and a test portion of the dataset, as taught by Wang. One would have been motivated to combine Wang with Sarferaz-Schmidt-Matlab-Clark, and would have had a reasonable expectation of success, as the combination saves a user time by having a pre-selected data split appropriate for a certain model.
Claim(s) 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz”, in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, as applied to Claim 1 above, and further in light of U.S. Patent Application Publication NO. 20210241177 “Wang”.
Claim 13:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz, Schmidt, and Matlab teach may not explicitly teach
wherein the at least one modification to a corresponding scenario setting includes a model-specific setting for nulls removal, a model-specific setting for outlier removal, a model-specific algorithm parameter, a model-specific data split configuration
However, Wang teaches wherein the at least one modification to a corresponding scenario setting includes a model-specific setting for nulls removal, a model-specific setting for outlier removal, a model-specific algorithm parameter, a model-specific data split configuration (i.e. para. [0119], “the configuration of automatic data splitting (training/validation) is also provided, the user may select “splitting by proportion”, “splitting by rule” and “sorting firstly and then splitting data”, and the proportion of the training set is further provided, the user may set the proportion to “0.8” and so on”, wherein a user may in real time update the specific model with a different ratio), or a model-specific filter condition that is to be added to a filter condition of the predictive scenario.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the at least one modification to a corresponding scenario setting includes a model-specific setting for nulls removal, a model-specific setting for outlier removal, a model-specific algorithm parameter, a model-specific data split configuration, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with a user may modify a model in real time by changing the automatic data splitting ratios, as taught by Wang. One would have been motivated to combine Wang with Sarferaz-Schmidt-Matlab, and would have had a reasonable expectation of success, as the combination provides a user with more granularity in achieving a desirable outcome.
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz” in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, as applied to Claim 1 above, in light of U.S. Patent Application Publication NO. 20160267396 “Gray” and further in light of U.S. Patent Application Publication NO. 20210241177 “Wang”.
Claim 15:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz, Schmidt, and Matlab may not explicitly teach
wherein generating the prepared dataset comprises pre-processing the dataset based on data pre-processing settings in the first model, filtering the dataset based on filter conditions in the first model parameters, and splitting the copy of the dataset into a training portion and a test portion based on a data split configuration in the first model parameters.
However, Gray teaches
wherein generating the prepared dataset comprises pre-processing the dataset based on data pre-processing settings in the first model , filtering the dataset based on filter conditions in the first model parameters (i.e. para. [0068], “data preprocessing may include data cleaning, removal of outliers, identifying and treating missing values, and transformation of values, etc”, wherein the BRI for filter conditions encompasses removing outliers in a selected dataset),
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein generating the prepared dataset comprises pre-processing the dataset based on data pre-processing settings in the first model , filtering the dataset based on filter conditions in the first model parameters, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with wherein generating the prepared dataset comprises pre-processing the dataset based on data pre-processing settings in the first model , filtering the dataset based on filter conditions in the first model parameters, as taught by Gray. One would have been motivated to combine Gray with Sarferaz-Schmidt-Matlab, and would have had a reasonable expectation of success, as the combination saves a user time and helps create a more accurate model.
While Sarferaz, Schmidt, Matlab, and Gray teach preprocessing data based on settings in a first selected model’s dataset, Sarferaz, Schmidt, Matlab, and Gray may not explicitly teach
splitting the copy of the dataset into a training portion and a test portion based on a data split configuration in the first model parameters.
However, Wang teaches
splitting the copy of the dataset into a training portion and a test portion based on a data split configuration in the first model parameters (i.e. para. [0046-0047], “the automatic machine learning technology may relate to at least one of: an automatic data splitting for splitting the historical data into training data and verification data… Specifically, the historical data may be automatically split into the training data and the verification data according to a preset splitting rule”, wherein dataset’s data is into at least a training a verification portion).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add splitting the copy of the dataset into a training portion and a test portion based on a data split configuration in the first model parameters, to Sarferaz-Schmidt-Matlab-Gray’s user construction and training of specified predictive models, with splitting the copy of the dataset into a training portion and a test portion based on a data split configuration in the first model parameters, as taught by Wang. One would have been motivated to combine Wang with Sarferaz-Schmidt-Matlab-Gray, and would have had a reasonable expectation of success, as the combination saves a user time by having a pre-selected data split appropriate for a certain model.
Claim(s) 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz”, in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, as applied to Claim 1 above, and further in light of U.S. Patent Application Publication NO. 20210241177 “Turco”.
Claim 16:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz, Schmidt, and Matlab may not explicitly teach
wherein the request to activate the first model is based on a comparison of the first model evaluation data to model evaluation data of at least one other model of the predictive scenario
However, Turco teaches,
wherein the request to activate the first model is based on a comparison of the first model evaluation data to model evaluation data of at least one other model of the predictive scenario (i.e. para. [0093], “For example, the plurality of models can be created by training a machine learning logic on training data comprising examples of good and bad machine translations from a source language (e.g. English) into one target language (e.g. French). … The best-model is the one of the plurality of models which (according to some KPI or evaluation metric) provides the most accurate assessment of the quality of a machine translation by predicting the correct effort”, wherein a model is chosen based on a comparison to find a best model).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add wherein the request to activate the first model is based on a comparison of the first model evaluation data to model evaluation data of at least one other model of the predictive scenario, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with wherein the request to activate the first model is based on a comparison of the first model evaluation data to model evaluation data of at least one other model of the predictive scenario, as taught by Turco. One would have been motivated to combine Turco with Sarferaz-Schmidt-Matlab, and would have had a reasonable expectation of success as it saves a user time by increasing the likelihood of finding the best model for the circumstance.
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application Publication NO. 20210209501 “Sarferaz”, in light of U.S. Patent Application Publication NO. 20180114135 “Schmidt-Karaca”, hereinafter “Schmidt”, and further in light of MATLAB. (2019, October 4). Choosing a Machine Learning Algorithm- MATLAB Live! YouTube. https://www.youtube.com/watch?v=VgyYWv908Y4, hereinafter “Matlab”, as applied to Claim 1 above, and further in light of U.S. Patent Application Publication NO. 20230306288 “Van Der Stockt”.
Claim 18:
Sarferaz, Schmidt, and Matlab teach the computer-implemented method of claim 1.
Sarferaz and Schmidt teach may not explicitly teach further comprising
receiving, from the predictive analytics library for a first prediction, field contribution data that indicates which fields of the dataset most contributed to the first prediction.
However, Van Der Stockt teaches
receiving, from the predictive analytics library for a first prediction, field contribution data that indicates which fields of the dataset most contributed to the first prediction (i.e. para. [0098], Model 620 is a machine learning model producing predictions 630 from transactions 610. Module 410 uses feature importance data produced by the explainability analysis to construct a superset of the most important features, within the input data of a transaction, in explaining predictions 610).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to add receiving, from the predictive analytics library for a first prediction, field contribution data that indicates which fields of the dataset most contributed to the first prediction, to Sarferaz-Schmidt-Matlab’s user construction and training of specified predictive models, with receiving, from the predictive analytics library for a first prediction, field contribution data that indicates which fields of the dataset most contributed to the first prediction, as taught by Van Der Stockt. One would have been motivated to combine Van Der Stockt with Sarferaz-Schmidt-Matlab, and would have had a reasonable expectation of success as the combination provides model explainability which is important in promoting user trust in the model's results, and helps those affected by a decision to challenge or change that outcome.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
U.S. Patent Application Publication NO. 20150379429 “Lee”, teaches in para. [0078], expert users may customize the parameters or settings they wish to use for various types of machine learning tasks, such as input record handling, feature processing, model building, execution and evaluation. In at least some embodiments, in addition to or instead of using pre-defined libraries implementing various types of machine learning tasks,.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DAVID H TAN whose telephone number is (571)272-7433. The examiner can normally be reached M-F 7:30-4:30.
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 at (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.
/D.T./Examiner, Art Unit 2145
/CESAR B PAULA/ Supervisory Patent Examiner, Art Unit 2145
