DETAILED ACTION
This final rejection is responsive to the amendment filed on December 11, 2025. Claims 1-2, 4-13, 15-17, 35, and 52 are pending. Claim 1 is independent. Claims 3 and 14 are cancelled.
Claim objections to claims 14-17 are withdrawn in light of applicant’s amendment.
Claim rejections under 35 USC §112 of claims 3-11 are withdrawn in light of applicant’s amendment.
Claim rejections under 35 USC §101 of claims 5-11 are withdrawn in light of applicant’s amendment. See section Response to Arguments below.
Claim rejections under 35 USC §102 and 103 of claims 1-2, 4-13, 15-17, 35, and 52 are maintained and merely updated to reflect the amended claim language. See sections Claim Rejections – 35 USC §102, Claim Rejections – 35 USC §103, and Response to Arguments below.
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on December 16, 2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (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.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-2, 4-5 and 15 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Hall et al. (The WEKA data mining software: An update), hereinafter Hall.
Regarding claim 1, Hall teaches the method for executing an automated machine learning process:
providing the GUI and providing in the GUI separately a model training operator and a model prediction operator that are mutually independent; (Hall, section 2, paragraph 2: “The workbench includes algorithms for regression, classification, clustering, association rule mining and attribute selection. Preliminary exploration of data is well catered for by data visualization facilities and many preprocessing tools. These, when combined with statistical evaluation of learning schemes and visualization of the results of learning, supports process models of data mining such as CRISP-DM [27].” – See also figure 6: The Explorer and KnowledgeFlow build models while the Experimenter compares models. These are separate applications within the WEKA workbench, which is a GUI, and are thus mutually independent.)
training a machine learning model on the basis of training data using the model training operator; and (Hall, section 2.1, page 11, col. 2, paragraph 4: “The Explorer is designed for batch-based data processing: training data is loaded into memory in its entirety and then processed. This may not be suitable for problems involving large datasets. However, WEKA does have implementations of some algorithms that allow incremental model building, which can be applied in incremental mode from a command-line interface.”)
providing a prediction service on prediction data using the model prediction operator and the trained machine learning model (Hall, section 2.1, page 11, col. 2, paragraph 5: “The third main graphical user interface in WEKA is the “Experimenter” (see Figure 3). This interface is designed to facilitate experimental comparison of the predictive performance of algorithms based on the many different evaluation criteria that are available in WEKA.” – The Experimenter is analogous to the prediction service as it allows for generating and training predictive models for comparison.)
wherein at least one of the following conditions is satisfied:
(1) said training of the machine learning model on the basis of training data using the model training operator comprises: providing a configuration interface for configuring model training in response to triggering operation on the model training operator; (Hall, section 2.1, paragraph 1: “The main graphical user interface is the “Explorer”. It has a panel-based interface, where different panels correspond to different data mining tasks. In the first panel, called “Preprocess” panel, data can be loaded and transformed using WEKA’s data preprocessing tools, called “filters”. This panel is shown in Figure 1.” – The preprocessing includes being able to select which attributes are being selected for the model training operator.) obtaining training samples by data preprocessing and feature engineering processing on the training data according to configuration information input by a user through the configuration interface; and (Hall, figure 1 – This shows that the training data is able to be loaded, generated, or edited – the data preprocessing – and then the desired attributes can be selected in order to obtain the samples by feature engineering based on the configuration input through the user interface.) training the machine learning model on the basis of the training samples using at least one model training algorithm. (Hall, Section 2.1, page 11, column 1, paragraph 1: “The second panel in the Explorer gives access to WEKA’s classification and regression algorithms. The corresponding panel is called “Classify” because regression techniques are viewed as predictors of “continuous classes”. By default, the panel runs a cross-validation for a selected learning algorithm on the dataset that has been been prepared in the Preprocess panel to estimate predictive performance. It also shows a textual representation of the model built from the full dataset.” – The training samples are acquired from the preprocessing tab of the Explorer which is then used to train a model using classification and regression algorithms.)
(2) said providing of the prediction service on prediction data using the model prediction operator and the trained machine learning model comprises: providing a configuration interface for configuring a batch prediction service in response to the triggering operation of the model prediction operator; (Hall, section 2.1, paragraph 1: “The main graphical user interface is the “Explorer”. It has a panel-based interface, where different panels correspond to different data mining tasks. In the first panel, called “Preprocess” panel, data can be loaded and transformed using WEKA’s data preprocessing tools, called “filters”. This panel is shown in Figure 1.” And Hall, section 2.1, page 11, column 2, paragraph 4: “The Explorer is designed for batch-based data processing:” – The Explorer, which is designed for batch-based processing, is therefore analogous to the batch processing service. The preprocessing includes being able to select which attributes are being selected for the model training operator.) obtain prediction samples by data preprocessing and feature-update processing on the training data using the model prediction operator according to configuration information input by a user through the configuration interface; and (Hall, figure 1 – This shows that the data is able to be loaded, generated, or edited – the data preprocessing – and then the desired attributes can be selected in order to obtain the samples by feature engineering based on the configuration input through the user interface.) providing prediction results for the prediction samples using the trained machine learning model. (Hall, figure 9 – The “Experiment output” panel includes prediction results for the samples. For instance, it includes the number of correct and incorrect as well as percentages, etc.)
Regarding claim 2, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
providing an editing interface according to operation of editing the model training operator; (Hall, figures 1, 3, 5, and 8-10 – Each of these figures shows an interface that is capable of editing the model training operator. Figure 8 shows a particular example of editing a Bayesian network for training.)
acquiring operator content input through the editing interface, wherein the operator content includes an operation command of data preprocessing on input training data, an operation command of feature engineering on the training data that has undergone data preprocessing, and an operation command of model training according to the results of feature engineering; and (Hall, section 2.1, paragraph 1: “The main graphical user interface is the “Explorer”. It has a panel-based interface, where different panels correspond to different data mining tasks. In the first panel, called “Preprocess” panel, data can be loaded and transformed using WEKA’s data preprocessing tools, called “filters”. This panel is shown in Figure 1. Data can be loaded from various sources, including files, URLs and databases. Supported file formats include WEKA’s own ARFF format, CSV, LibSVM’s format, and C4.5’s format. It is also possible to generate data using an artificial data source and edit data manually using a dataset editor.” – In addition to the preprocessing and feature engineering that occurs in the Preprocess panel with the ability to select filter and attributes, figure 1 also shows a “classify” and “cluster” tab which are models that can be trained using the training data thus, this is analogous to acquiring the content input through the editing interface.)
encapsulating the operator content to obtain the model training operator. (Hall, section 2.1, page 11, column 1, paragraph 1: “By default, the panel runs a cross-validation for a selected learning algorithm on the dataset that has been been prepared in the Preprocess panel to estimate predictive performance.” – The training data that was prepared and then used in cross-validation is analogous to encapsulating the operator content.)
Regarding claim 4, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
the configuration interface includes at least one of the following configuration items:
input source configuration item of a machine learning model; (Hall, figure 1 – The ability to “open file…”, “open URL…”, or “Open DB…” is an input source configuration item.)
applicable problem type configuration item of a machine learning model; (Hall, figure 3, “Experiment Type” – The choice between classification and regression is analogous to the applicable problem type as the specification defines the applicable problem type to be “any one of a binary classification problem, a regression problem and a multi classification problem” in paragraph 0057.)
algorithm mode configuration item for training a machine learning model;
optimization objective configuration item of a machine learning model; and
field name configuration item of a prediction objective field of a machine learning model.
Regarding claim 5, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
wherein the data preprocessing on the training data comprises at least one of the following items:
Item 1, performing data type conversion of the training data; (Hall, section 4.3, bullet 8 – Converts numeric to nominal which is data type conversion.)
Item 2, sampling the training data; (Hall, section 4.3, bullet 11 – selecting a random subset is analogous to sampling.)
Item 3, annotating the training data as labeled data and unlabeled data;
Item 4, automatically identifying a data field type of the training data;
Item 5, filling in missing values of the training data;
Item 6, analyzing an initial time field of the training data, obtaining and adding a new time field, and deleting the initial time field;
Item 7, automatically identifying non-numerical data in the training data, and hashing the non-numerical data.
Regarding claim 15, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
the configuration interface comprises at least one of:
a configuration item of field selection in a prediction result, and (Hall, Figure 9 – “Result list” states to right-click for options and therefore provides configuration item of field selection in a prediction result with the options being the configurations.)
a configuration item of switching state of a simulated real-time prediction service.
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.
Claim(s) 6, 7, 9, 12, 13, 35, and 52 is/are rejected under 35 U.S.C. 103 as being unpatentable over Hall in view of Bansal et al. (US20210097444), hereinafter Bansal.
Regarding claim 6, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
sampling the training data that has undergone data preprocessing; (Hall, figure 3 – The “Experiment Type” panel is capable of performing cross validation where the data is divided into subsets which is indicative that the data that was preprocessed in claim 3 is then sampled.)
Hall does not explicitly teach:
performing feature pre-selection on the training data that has undergone the sampling, to obtain basic features;
performing feature derivation on the basic features to obtain derived features; and
generating training samples according to the basic features and the derived features.
However, Bansal teaches:
performing feature pre-selection on the training data that has undergone the sampling, to obtain basic features; performing feature derivation on the basic features to obtain derived features; and generating training samples according to the basic features and the derived features. (Bansal, paragraph 0054: “Thereafter, the actual preprocessing can be performed by a set of feature preprocessors 515A-515N, where each preprocessor may implement a particular preprocessing step, or may implement multiple preprocessing steps (e.g., for one pipeline). The feature preprocessors 515A-515N may thus use the data from the feature preprocessing analyzers to apply the preprocessing operations/transforms to the dataset (e.g., a row at a time), optionally at least partially in parallel, to yield a transformed output dataset (or portion thereof).” – Figure 8 additionally shows the different preprocessing pipelines with one being a “baseline” feature preprocessing. Thus, the baseline is analogous to the basic features being pre-selected. The different transformations is analogous to performing feature derivation and the output dataset is then analogous to training samples according to the basic features and the derived features.)
Bansal is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall, which already teaches obtaining training samples by data preprocessing and feature engineering but does not explicitly teach that the data preprocessing and feature engineering includes obtaining basic features and derived features and generating training samples from those, to include the teachings of Bansal which does teach that the data preprocessing and feature engineering includes obtaining basic features and derived features and generating training samples from those in order to make it easier for those who don’t have the required knowledge to “clean up the data or preprocess it in order to build good models.” (Bansal, paragraph 0002)
Regarding claim 7, Hall and Bansal teach the method of claim 6, as cited above.
Hall does not explicitly teach:
extracting all attribute information included in the training data that has undergone the sampling, wherein the attribute information is used to form features;
acquiring feature importance values of each attribute information; and
obtaining the basic features according to the feature importance values.
However, Bansal further teaches:
extracting all attribute information included in the training data that has undergone the sampling, wherein the attribute information is used to form features; (Bansal, paragraph 0047: “The preliminary operations may also include generating metadata describing the dataset (e.g., a total number of rows, a number of columns, data types of the columns, value distributions and other statistics based on values of the columns) that can be used as part of later processing, cleaning the dataset, or the like. – The metadata that is being generated is analogous to extracting the attribute information; while the metadata being used as part of later processing would include being used to form features – see below.)
acquiring feature importance values of each attribute information; and (Bansal, paragraph 0044: “The pipeline recommender system 112, in some embodiments, can analyze the user's provided dataset and infer one or more of the probabilistic schema of the data set, target leakage, feature importance, the type of ML problem (e.g., classification, regression, etc.) based on the user-identified target column, etc.” – The pipeline recommender system inferring the feature importance of the dataset is analogous to acquiring feature importance values of each attribute information.)
obtaining the basic features according to the feature importance values. (Bansal, paragraph 0044: “The pipeline recommender system 112 can use the dataset provided by the user and knowledge learned from metadata collected to recommend a promising and diverse set of feature processing pipelines to apply to the customer dataset along with the code for the feature processing model.” – The feature processing applied to the dataset is analogous to obtaining the basic features according to the feature importance values as that is part of the metadata collected and used to create the features.)
Regarding claim 9, Hall and Bansal teach the method of claim 6, as cited above.
Hall does not explicitly teach:
performing at least one of statistical calculation and feature combination on the basic features to obtain the derived features, using preset feature generation rules.
However, Bansal further teaches:
performing at least one of statistical calculation and feature combination on the basic features to obtain the derived features, using preset feature generation rules. (Bansal, paragraph 0050: “As one example, the pipeline recommender system 112 may recommend up to ten pipelines to explore, such as (1) “apply one-hot encoding and principal component analysis (as the feature preprocessors/transforms) followed by use of the ‘XGBOOST’ algorithm with hyperparameter tuning”, (2) “apply one-hot encoding and principal component analysis (as the feature preprocessors/transforms) followed by use of a ‘linear learner’ algorithm with hyperparameter tuning”, (3) “apply principal component analysis (as the feature preprocessor/transform) followed by use of the ‘XGBOOST’ algorithm with hyperparameter tuning”, and the like.” – one-hot encoding, principal component analysis, and XGBOOST are all examples of feature generation rules. One-hot encoding and principal component analysis are both methods of feature combination while XGBOOST incorporates the L1 or L2 norm and thus involves a statistical calculation.)
Regarding claim 12, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
obtaining a model training scheme on the basis of the trained machine learning model; and (Hall, Figure 9 – The experiment tab has an “experiment output” section that gives information of the trained model.)
visualizing the model training scheme; (Hall, section 4.5, paragraph 2: “Similar mechanisms allow new visualizations for classifier errors, predictions, trees and graphs to be added to the pop-up menu available in the history list of the Explorer’s “Classify” panel.”)
wherein the model training scheme includes any one or more of: an algorithm used to train the machine learning model, (Hall, figure 5 – Gives information about the algorithm used to train the model.) hyperparameters of the machine learning model, effects of the machine learning model, (Hall, figure 9 – The specification states that the hyperparameters are model hyperparameters and training hyperparameters and that the training hyperparameters are “learning rate, batch size, and number of iterations” in paragraph 00167. The “Experiment output” of figure 5 shows that this includes the number of runs which is analogous to the number of iterations. In addition, the effects of the machine learning model would include the information provided under the “measure” heading in the “Experiment output” of figure 5.)
Hall does not explicitly teach:
That the model training scheme includes feature information.
wherein the feature information includes any one or more of feature quantity, feature generation method and feature importance analysis results.
However, Bansal teaches:
That the model training scheme includes feature information. (Bansal, figure 8 – The “define feature processing pipelines” panel gives feature information.)
wherein the feature information includes any one or more of feature quantity, feature generation method and feature importance analysis results. (Bansal, paragraph 0066: “As shown, a first pipeline “FP_BASELINE” is defined with a first “FPO” step, which is defined (as a training job to learn the transformations) with values for a source directory, instance types and counts, and other non-illustrated values such as an ML framework version to be used, a set of tags to be applied, an identifier of a feature processing strategy (e.g., a baseline strategy that performs a 1-hot encoding of all categorical variables and does a median-impute null values with indicators;” – This is describing the feature processing pipelines in figure 8, the feature processing strategy is analogous to the feature generation method.)
Bansal is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall, which already teaches training a model, obtaining a model training scheme and visualizing the model training scheme but does not explicitly teach that the model training scheme includes feature information of a feature quantity, feature generation method, or feature importance analysis results, to include the teachings of Bansal which does teach that the model training scheme includes feature information of a feature quantity, feature generation method, or feature importance analysis results in order to make it easier for those who don’t have the required knowledge to “clean up the data or preprocess it in order to build good models.” (Bansal, paragraph 0002)
Regarding claim 13, Hall and Bansal teach the method of claim 12, as cited above.
Hall further teaches:
a step of retraining the machine learning model according to the preview results of the visualization. (Hall, section 2.1, page 11, column 2, paragraph 3: “However, in addition to batch-based training, its data flow model enables incremental updates with processing nodes that can load and preprocess individual instances before feeding them into appropriate incremental learning algorithms. It also provides nodes for visualization and evaluation. Once a set-up of interconnected processing nodes has been configured, it can be saved for later re-use.” – The incremental learning is analogous to retraining the model according to the results as the nodes for visualization are a part of the incremental learning process which means that each iteration will receive the visualization results and then more training is done based off of those results.)
Regarding claim 35, Hall teaches the method of claim 1, as cited above.
Hall does not explicitly teach:
An apparatus comprising at least one computing device and at least one storage device, wherein the at least one storage device is configured to store instructions, wherein the instructions are configured, upon being executed by the at least one computing device, to cause the at least one computing device to execute the method of claim 1 for execution of an automated machine learning process.
However, Bansal teaches:
An apparatus comprising at least one computing device and at least one storage device, wherein the at least one storage device is configured to store instructions, wherein the instructions are configured, upon being executed by the at least one computing device, to cause the at least one computing device to execute the method of claim 1 for execution of an automated machine learning process. (Bansal, paragraph 0018: “The present disclosure relates to methods, apparatus, systems, and non-transitory computer-readable storage media for automated machine learning pipeline exploration and deployment”, paragraph 0151: “Various embodiments discussed or suggested herein can be implemented in a wide variety of operating environments, which in some cases can include one or more user computers, computing devices, or processing devices which can be used to operate any of a number of applications.” and paragraph 0154: “Such a system may also include one or more storage devices, such as disk drives, optical storage devices, and solid-state storage devices such as random-access memory (RAM) or read-only memory (ROM), as well as removable media devices, memory cards, flash cards, etc.”)
Bansal is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall, which already teaches a method for execution of an automated learning process but does not explicitly teach an apparatus with at least one computing device and at least one storage device where the instructions of the method are stored and executed, to include the teachings of Bansal which does teach an apparatus with at least one computing device and at least one storage device where the instructions of the method are stored and executed. Although Hall states that WEKA is a software (Hall, abstract) it does not specify where it is stored and executed, whereas Bansal states in paragraph 0020: “In FIG. 1, an automated machine learning pipeline generation system 102 (or “AMPGS”) implemented as part of a machine learning service 110 develops, evaluates, and/or deploys ML pipelines on behalf of users 109. The AMPGS 102 (and ML service 110) may be implemented as software, hardware, or a combination of both using one or more computing devices in one or multiple networks and/or geographic locations.” Thus, combining the software of Lee with the hardware components of Bansal’s machine learning service would yield the predictable result of storing executing the software instructions to automate machine learning.
Regarding claim 52, Hall teaches the method of claim 1, as cited above.
Hall does not explicitly teach:
A computer-readable storage medium having a computer program stored thereon, which computer program, when executed by a processor, implements the method of claim 1.
However, Bansal teaches:
A computer-readable storage medium having a computer program stored thereon, which computer program, when executed by a processor, implements the method of claim 1. (Bansal, paragraph 0018: “The present disclosure relates to methods, apparatus, systems, and non-transitory computer-readable storage media for automated machine learning pipeline exploration and deployment”)
Bansal is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall, which already teaches a method for execution of an automated learning process but does not explicitly teach a computer-readable storage medium where the computer program is stored and executed by a processor, to include the teachings of Bansal which does teach computer-readable storage medium where the computer program is stored and executed by a processor. Although Hall states that WEKA is a software (Hall, abstract) it does not specify where it is stored and executed, whereas Bansal states in paragraph 0020: “In FIG. 1, an automated machine learning pipeline generation system 102 (or “AMPGS”) implemented as part of a machine learning service 110 develops, evaluates, and/or deploys ML pipelines on behalf of users 109. The AMPGS 102 (and ML service 110) may be implemented as software, hardware, or a combination of both using one or more computing devices in one or multiple networks and/or geographic locations.” Thus, combining the software of Lee with the hardware components of Bansal’s machine learning service would yield the predictable result of storing executing the software instructions to automate machine learning.
Claim(s) 8, 10, and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Hall in view of Bansal in view of Ved (Feature Selection in Machine Learning: Variable Ranking and Feature Subset Selection Methods), hereinafter Ved.
Regarding claim 8, Hall and Bansal teach the method of claim 7, as cited above.
Hall and Bansal do not explicitly teach:
wherein said obtaining the basic features according to the feature importance values comprises:
ranking all the feature importance values to obtain a ranking result; and
acquiring a first predetermined quantity of attribute information as the basic features according to the ranking result.
However, Ved teaches:
wherein said obtaining the basic features according to the feature importance values comprises:
ranking all the feature importance values to obtain a ranking result; and (Ved, Page 2, last 2 paragraphs: “Variable Ranking is the process of ordering the features by the value of some scoring function, which usually measures feature-relevance. Resulting set: The score S(fi) is computed from the training data, measuring some criteria of feature fi. By convention a high score is indicative for a valuable (relevant) feature.” – The basic features obtained in claim 7 by Hall and Bansal are analogous to the training data which is then used to score (rank) the features. These scores are analogous to the ranking result.)
acquiring a first predetermined quantity of attribute information as the basic features according to the ranking result. (Ved, page 3, paragraph 1: “A simple method for feature selection using variable ranking is to select the k highest ranked features according to S. This is usually not optimal, but often preferable to other, more complicated methods. It is computationally efficient — only calculation and sorting of n scores.” – the k highest ranked features is indicative that k is a predetermined quantity.)
Ved is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall and Bansal, which already teaches obtaining basic features according to feature importance but does not explicitly teach ranking the feature importance values to obtain a ranking result and acquiring a predetermined number of attributes based on that, to include the teachings of Ved which does teach ranking the feature importance values to obtain a ranking result and acquiring a predetermined number of attributes based on that in order to “select variables according to their individual predictive power.” (Ved, page 6, last paragraph)
Regarding claim 10, Hall and Bansal teach the method of claim 6, as cited above.
Hall and Bansal do not explicitly teach:
performing feature post-selection on the basic features and the derived features; and
generating training samples according to features obtained through the feature post-selection.
However, Ved teaches:
performing feature post-selection on the basic features and the derived features; and (Ved, page 7, paragraph 2: “For Feature Subset Selection you’d need: A measure for assessing the goodness of a feature subset (scoring function)” – The specification states that feature post-selection includes “acquiring feature importance values of each basic feature and each derived feature” in paragraph 00104. Therefore, feature subset selection which includes a scoring function is analogous to feature post-selection, given the set of basic features and derived features from claim 6.)
generating training samples according to features obtained through the feature post-selection. (Ved, page 7, paragraph 1: “The Goal of Feature Subset Selection is to find the optimal feature subset.” – The optimal feature subset is analogous to the training samples as the subset is being generated based off the feature post-selection.)
Ved is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall and Bansal, which already teaches obtaining derived features but does not explicitly teach performing post-selection on the basic and derived features and generating training samples according to that, to include the teachings of Ved which does teach performing post-selection on the basic and derived features and generating training samples since it provides “generic selection of features, not tuned by given learner (universal).” (Ved, page 8, fourth bullet point)
Regarding claim 11, Hall, Bansal, and Ved teach the method of claim 10, as cited above.
Hall and Bansal do not explicitly teach:
acquiring feature importance values of each basic feature and each derived feature, ranking all the feature importance values to obtain a ranking result; and
acquiring a second predetermined quantity of features as required features for generating training samples, according to the ranking result.
However, Ved further teaches:
acquiring feature importance values of each basic feature and each derived feature, ranking all the feature importance values to obtain a ranking result; and (Ved, Page 2, last 2 paragraphs: “Variable Ranking is the process of ordering the features by the value of some scoring function, which usually measures feature-relevance. Resulting set: The score S(fi) is computed from the training data, measuring some criteria of feature fi. By convention a high score is indicative for a valuable (relevant) feature.” – The basic and derived features obtained in claim 6 by Hall and Bansal are analogous to the training data which is then used to score (rank) the features. These scores are analogous to the ranking result.)
acquiring a second predetermined quantity of features as required features for generating training samples, according to the ranking result. (Ved, page 3, paragraph 1: “A simple method for feature selection using variable ranking is to select the k highest ranked features according to S. This is usually not optimal, but often preferable to other, more complicated methods. It is computationally efficient — only calculation and sorting of n scores.” – the k highest ranked features is indicative that k is a predetermined quantity.)
Claim(s) 16 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Hall in view of Bansal in view of Thandu (Capabilities of Google’s Prediction API for building practical machine-learning based applications), hereinafter Thandu.
Regarding claim 16, Hall teaches the method of claim 1, as cited above.
Hall further teaches:
providing a configuration interface for configuring a real-time prediction service according to an operation of configuring a real-time prediction service; (Hall, section 2.1, page 11, column 2, paragraph 4: “Most tasks that can be tackled with the Explorer can also be handled by the Knowledge Flow. However, in addition to batch-based training, its data flow model enables incremental updates with processing nodes that can load and preprocess individual instances before feeding them into appropriate incremental learning algorithms.” – The specification’s discussion of “a real-time prediction service” is merely a discussion of whether the switching state is “on” or “off” – see paragraph 00124. When it is off then it is a batch prediction and “the prediction results of each sample may influence each other” and when it is on the “prediction samples will not influence each other.” Using the Knowledge Flow interface instead of the Explorer interface allows the individual instances to be fed into the algorithms which therefore ensures that they don’t influence each other. Figure 2 shows the Knowledge Flow interface with the tabs at the top including many of the same functionality as the Explorer, e.g. the Filters, which shows that it is capable of configuring in the same way as presented in claim 14.)
Hall does not explicitly teach:
receiving a prediction service request including the prediction data through the API address provided in the configuration interface; and
obtaining a prediction result on the prediction data in response to the received prediction service request using the selected machine learning model, and sending the prediction result through the API address.
However, Thandu teaches:
receiving a prediction service request including the prediction data through the API address provided in the configuration interface; and (Thandu, page 5, paragraph 4 through page 6, paragraph 2: “To use the Prediction API, the only cloud service required is “Cloud Storage,” which is enabled by default in your Google Cloud Platform project. You are required to create a bucket in the location where your training set is uploaded. The Prediction API offers a simple way to train machine learning models through aRESTful interface. To authorize the requests, your application must use “OAuth 2.0” protocol; the API does not support any other types of authorization protocol. The application wrapping this model could use Google Sign-In for some aspects of authorization to the API. The detailed information on the authorization is given in GoogleOAuth (https://developers.google.com/identity/protocols/OAuth2) documentation.” – The application (configuration interface) using Google Sign-In is indicative that the API address is provided in the configuration interface.)
obtaining a prediction result on the prediction data in response to the received prediction service request using the selected machine learning model, and sending the prediction result through the API address. (Thandu, page 6, paragraph 7: “For the resulting model that the API builds, you can make predictions for new example datasets by calling trainedmodels.predict method. This returns the parameters outputLabel (numeric or String) and outputMulti which provide probability measures for each prediction class.” – The prediction outputs that are provided is analogous to obtaining a prediction result. This is sent through the application and is therefore sent through the API address.)
Thandu is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Hall, which already teaches a prediction service for obtaining prediction results but does not explicitly teach receiving the prediction data and results through the API address, to include the teachings of Thandu which does teach receiving the prediction data and results through the API address in order to “access many advanced machine-learning capabilities through fast, reliable, and cost-effective prediction API infrastructures.” (Thandu, page 1, paragraph 2)
Regarding claim 17, Hall and Thandu teach the method of claim 16, as cited above.
Hall further teaches:
the configuration interface comprises at least one of:
a configuration item for model selection rules for selecting an online machine learning model from the trained machine learning models, and (Hall, figure 2 – shows a number of different models figure 3 – “algorithms” allows for the selection of machine learning models from the trained models)
a configuration item for application resources.
Response to Arguments
Applicant’s arguments, see 8-9 of Applicant's Remarks, filed December 11, 2025, with respect to claim rejections of claims 5-11 under 35 USC §101 have been fully considered and are persuasive. In particular, paragraph 1 of page 9 states that the abstract idea is not being performed in isolation but is part of specific ordered steps that provides a technological solution. The rejection of claims 5-11 under 35 USC §101 has been withdrawn.
Applicant's arguments filed December 11, 2025 regarding claim rejections under 35 USC §102 and §103 have been fully considered but they are not persuasive. Applicant’s arguments appear to rely on language recited in preamble recitations in claim 1. When reading the preamble in the context of the entire claim, the recitation "an automated machine learning process" is not limiting because the body of the claim describes a complete invention and the language recited solely in the preamble does not provide any distinct definition of any of the claimed invention’s limitations. Thus, the preamble of the claim(s) is not considered a limitation and is of no significance to claim construction. See Pitney Bowes, Inc. v. Hewlett-Packard Co., 182 F.3d 1298, 1305, 51 USPQ2d 1161, 1165 (Fed. Cir. 1999). See MPEP § 2111.02.
In response to applicant's argument on page 10 that Hall fails to teach or suggest “a guided configuration interface triggered by operator interaction or of mutually independent, directly invocable “operators” within a GUI,” examiner notes that 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). Nothing in the claim language suggest that the claimed limitation are performed in a guided manner, as suggested by applicant’s arguments. Hall’s WEKA requiring manually navigating between panels/tabs does not preclude it from teaching the claimed limitations. The claims themselves do not specify that the steps provided are being guided by the GUI or performed in a single panel/tab. The simplified nature of the GUI as illustrated in the cited figures 3-6 is not represented within the claim language.
Additionally, applicant argues on page 11 that claim 1 recites “a model training operator” and a “model prediction operator” that are pre-packaged, fully-functional, independent service units is somehow different from Hall’s Explorer/KnowledgeFlow to build models and the Experimenter which allows for comparing models’ predictive power without providing details on how they are different. Each of these “applications” of Hall’s WEKA are provided within the toolbox and would thus be pre-packaged, fully-functional, independent units.
Regarding applicant’s arguments on page 11 that Hall does not disclose or suggest the features of claim 2, examiner disagrees. As noted in the claim rejections above, under the preprocess tab of the explorer interface the user is able to load, transform, filter, and generate data which is all a part of preprocessing and feature engineering. In addition to that, the select attributes tab allows the user to identify and select the most important attributes of the data (see Hall, page 11, top of column 2). While examiner notes that the capability for a user to save and “invoke a custom, encapsulated workflow as a first-class entity within a GUI” is not explicitly recited in the claim language this is a capability shown in Hall in the open/save/log features in at least figures 1, 2, 3, and 5.
Regarding Applicant’s arguments on page 12 that Hall and Bansal do not disclose or suggest the “specific, sequential workflow of claim 6,” examiner disagrees. As noted above, the claim limitations do not require the steps to be automated. Thus, a “manual, step-by-step execution of these tasks” is analogous to a specific, sequential workflow. Additionally, the features that are lacking from Hall are taught by Bansal. Paragraph 0054, which is relied upon to teach the limitations of claim 6, is describing the sequential workflow outlined in Fig. 5.
Regarding applicant’s arguments on page 13 that Hall and Bansal fail to teach the visualization of “a model training scheme,” examiner disagrees. The language of claim 12 does not require a “unified visualization” as argued by applicant as the claim recites “wherein the model training scheme includes any one or more of: an algorithm used to train the machine learning model, hyperparameters of the machine learning model, effects of the machine learning model and feature information; wherein the feature information includes any one or more of feature quantity, feature generation method and feature importance analysis results.” Thus, Hall teaching a visualization of the model training scheme including an algorithm used, hyperparameters of the model and effects of the machine learning model is, in itself, enough to teach the claimed limitation. Bansal teaches the visualization of the feature information and the motivation to combine the two references to arrive at all the recited limitations would be to make it easier for those who don’t have the required knowledge to visualize how the preprocessing helps to build the models, as noted in the rejection of claim 12 (Hall, paragraph 0002).
Regarding applicant’s argument on page 13 that Hall and Bansal do not teach the limitations of claim 13, examiner disagrees. Applicant argues that “the step of ‘retraining the machine learning model according to the preview results of the visualization’, as recited in claim 13, creates a direct interactive feedback loop”, however, nothing in the recited limitation requires any user interaction. Retraining based on the preview results of the visualization does not require a user to provide any feedback. Thus, Hall’s system of visualizing the nodes of the incremental training at each iteration is analogous to retraining according to the preview results of the visualization.
Regarding applicant’s argument on page 14 that Hall, Bansal, and Thandu do not teach the limitations of claim 16, examiner disagrees. Though not relied upon for the claim rejection, Hall does discuss an integrated API on page 10, section 2. Thus, the more specific prediction API taught by Hall and relied upon for the limitations of “receiving a prediction service request…” and “obtaining a prediction result…” would be integrated into the service disclosed by Hall, rather than integrating the service into the API of Thandu.
Therefore, claim rejections of claims 1-2, 4-5, and 15 under 35 USC §102 and of claims 6-13, 16-17, 35, and 52 under 35 USC §103 are maintained. See sections Claim Rejections – 35 USC §102 and Claim Rejections – 35 USC §103 above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Bisson-Krol et al. (US20210110298A1) provides a method of pre-processing data, training a model, making predictions and editing via a user interface.
Dirac et al. (EP4328816) provides a machine learning services which uses programmatic interfaces for obtaining data sets and configuring feature processing and model creation.
Murakonda et al. (US20220351083) provides a GUI for configuring machine learning models and also includes interactive visualizations of the predictions.
Qi et al. (US20210304056) provides a method for configuring parameters for an automated machine learning process.
Squires et al. (US11443239) teaches a user interface for configuring input data and applying it for creating a machine learning model.
Hutter et al. (Automated Machine Learning: Methods, Systems, and Challenges) part II, starting on page 79, discusses AutoML Systems.
Dudley et al. (A Review of User Interface Design for Interactive Machine Learning)
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 JACQUELINE MEYER whose telephone number is (703)756-5676. The examiner can normally be reached M-F 8:00 am - 4:30 pm EST.
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, Tamara Kyle can be reached at 571-272-4241. 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.
/J.C.M./Examiner, Art Unit 2144
/TAMARA T KYLE/Supervisory Patent Examiner, Art Unit 2144