Automated Feature Graph Generation with Similarity Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models face reduced prediction performance due to the inclusion of unnecessary features in the data set, which increases computing resources and time, highlighting the need for an efficient method to automate feature engineering.
Innovation Solution
A computer-implemented method and apparatus for generating higher-level features by applying predefined augmentations to lower-level features, computing a bivariate similarity metric, and adding the new features to a feature graph only if the metric is below a threshold, thereby reducing redundant information and improving prediction performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual feature engineering is performed to generate useful new features, then prediction performance is improved, but time consumption and complexity increase
Solution Approach 1:
The system performs automated feature engineering where the computer automatically generates higher-level features from lower-level features using predefined augmentations, eliminating the need for manual expert intervention. The system serves itself by autonomously computing similarity metrics and selecting features based on thresholds, thereby reducing time consumption while maintaining prediction performance.
Solution Approach 2:
The system transforms features through predefined augmentations that apply parameter changes (transformations and combinations) to generate higher-level features. By systematically varying feature parameters through automated operations, the system efficiently explores feature space without manual intervention, improving prediction performance while reducing the time required compared to manual feature engineering.
2Reliability
If as many features as possible are generated to improve prediction, then more information is available, but computing resources and time increase
Solution Approach 1:
The system extracts only the most useful features by computing similarity metrics between generated higher-level features and existing lower-level features. Features with similarity above a predefined threshold are filtered out, extracting only those features that provide novel information. This selective extraction improves prediction performance while minimizing the use of computing resources by avoiding processing of redundant features.
Solution Approach 2:
Instead of generating all possible features (excessive action), the system generates features using predefined augmentations and selectively keeps only those below a similarity threshold (partial action). This partial approach generates sufficient useful features for good prediction performance without the computational burden of exhaustive feature generation, thereby optimizing the balance between prediction quality and resource consumption.
3Productivity
If feature selection is performed to reduce data, then computing efficiency is improved, but prediction performance may deteriorate
Solution Approach 1:
The system uses feedback through similarity metric computation to guide feature selection. By continuously evaluating the similarity between generated higher-level features and existing features, the system provides feedback on feature redundancy. This feedback mechanism ensures that only features adding novel information are selected, maintaining prediction performance while improving computing efficiency through reduced feature sets.
Solution Approach 2:
The system performs preliminary feature engineering by automatically generating higher-level features and filtering them through similarity thresholds before the main machine learning process. This preliminary action prepares an optimized feature set in advance, ensuring that only useful, non-redundant features are passed to the modeling stage, thereby maintaining prediction performance while improving subsequent computing efficiency.
Data Source
AI summary
A computer-implemented method for generating higher-level features based on one or more lower-level features of a data set includes generating a higher-level feature using a predefined augmentation of one or more lower-level features, wherein the predefined augmentation comprises a predefined transformation of a lower-level feature and/or a predefined combination of a plurality of lower-level features. The method further includes computing a bivariate similarity metric indicative of a similarity between the generated higher-level feature and the one or more lower-level features. Furthermore, the method comprises adding the higher-level feature to a feature graph, if the metric is less than a predefined threshold. Further, the method comprises outputting a result indicative of the feature graph comprising the lower-level features and the higher-level features.