Parallel Feature Selection via Rank Fusion for ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current feature selection methods for constructing machine learning models are labor-intensive and inefficient, often resulting in unstable and inaccurate feature selection due to reliance on single evaluation indices and limited consideration of comprehensive feature performance.
Innovation Solution
A method that divides a training data set into multiple subsets, processes each subset in parallel to compute evaluation indices and importance ranks, and fuses these ranks to obtain comprehensive total ranks for selecting target features, thereby considering the performance of features across multiple subsets and indices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual feature selection is performed, then feature selection accuracy can be improved through careful testing and filtering, but the process becomes very labor-intensive and slows down model construction speed
Solution Approach 1:
The system automatically performs feature selection by dividing training data into multiple subsets, computing evaluation indices for each subset, and fusing the results to select target features. This self-service mechanism eliminates manual feature selection operations while maintaining selection quality through multi-subset comprehensive evaluation.
2Productivity
If conventional automatic feature selection methods are used, then model construction speed is improved through automation, but feature selection accuracy deteriorates because only single evaluation indices and limited data subsets are considered
Solution Approach 1:
The training data set is segmented into multiple training data subsets, and feature evaluation is performed on each subset independently. This segmentation allows comprehensive evaluation of feature performance across different data portions, improving selection accuracy while maintaining automation through parallel processing of multiple subsets.
Solution Approach 2:
The system merges multiple evaluation indices computed on different training data subsets by fusing their results into total ranks. This combining approach integrates diverse evaluation perspectives to achieve more accurate and stable feature selection compared to single-index methods.
3Reliability
If multiple evaluation indices are computed and fused across multiple data subsets, then feature selection stability and accuracy are improved, but computational complexity increases
Solution Approach 1:
The patent replaces complex manual feature selection processes with an automated computational system that systematically divides data, computes multiple evaluation indices, and fuses results through rank aggregation. This substitution of mechanical/manual operations with automated algorithms manages computational complexity while improving reliability.
Data Source
AI summary
A feature-selection system obtains a training data set and associated features; divides the training data set into a first number of training data subsets; and forms a plurality of feature-selecting data sets. A feature-selecting data set comprises a second number of training data subsets. The system processes, in parallel, each feature-selecting data set, which comprises: computing a first evaluation index for the features based on the feature-selecting data set; obtaining a group of index ranks corresponding to the features based on the first evaluation index; and obtaining a group of importance ranks corresponding to the features based on the feature-selecting data set and a machine-learning model. The system further obtains a group of total ranks by fusing groups of index ranks and groups importance ranks obtained from processing the plurality of feature-selecting data sets; and selecting target features from the features based on the group of total ranks.