Feature Amount Selection via Principal Component Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-dimensional feature amount data analysis faces challenges in maintaining accuracy while reducing the number of feature amounts, leading to increased calculation time and decreased prediction accuracy due to the large number of explanatory variables.
Innovation Solution
A feature amount selection device that performs principal component analysis in a sample space and selects feature amounts based on the results, specifically choosing features with no distortion in distribution and a large distance from the origin, to speed up calculations without compromising analysis accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If the number of feature amounts is reduced to speed up calculation, then calculation time is shortened, but analysis accuracy may be deteriorated due to loss of information
Solution Approach 1:
The patent introduces an intermediary selection process that uses statistical measures (variance, correlation coefficients) as mediators to evaluate and select feature amounts. This intermediary evaluation mechanism allows the system to reduce the number of feature amounts while maintaining analysis accuracy by objectively identifying the most informative features without losing critical information.
Solution Approach 2:
The patent changes the parameters used for feature selection by employing multiple statistical measures (variance, correlation coefficients, etc.) to evaluate feature amounts. By changing from simple counting methods to multi-parameter statistical evaluation, the system can identify the most informative subset of features, thereby reducing dimensionality while preserving analysis accuracy.
2Measurement precision
If a wrapper method is used to select main feature amounts based on machine learning accuracy, then analysis accuracy is improved, but the number of combinations increases and calculation amount becomes enormous
Solution Approach 1:
The patent segments the feature selection process into two stages: first, statistical pre-screening using variance and correlation coefficients to eliminate obviously uninformative features; second, wrapper method evaluation only on the reduced subset. This segmentation dramatically reduces the combinatorial explosion while maintaining the accuracy benefits of wrapper methods.
Solution Approach 2:
The patent performs preliminary filtering of feature amounts using statistical measures (variance, correlation coefficients) before applying the computationally intensive wrapper method. This preliminary action reduces the search space for the wrapper method, making it feasible to apply to large-scale data while still achieving high analysis accuracy.
3Productivity
If a filter method is used to select feature amounts based on statistical numerical values, then calculation amount is reduced, but information obtained by fusing multiple feature amounts may be removed
Solution Approach 1:
The patent merges multiple statistical evaluation methods (variance-based filtering, correlation coefficient analysis, and wrapper method evaluation) into a comprehensive feature selection framework. By combining these methods, the system preserves fused information from multiple feature amounts while maintaining calculation efficiency through the hierarchical structure of the combined approach.
Solution Approach 2:
The patent performs preliminary statistical analysis (variance, correlation coefficients) to identify and eliminate obviously uninformative features before applying more sophisticated evaluation methods. This preliminary action preserves important fused information by ensuring that only truly uninformative features are removed, while maintaining calculation efficiency through the staged approach.
Data Source
AI summary
A feature amount selection device includes a feature amount data acquisition unit that acquires feature amount data including a set of values of a plurality of feature amounts for a sample, for each of a plurality of the samples, a principal component analysis unit that performs, on the feature amount data, principal component analysis in a sample space that is a collection of the plurality of feature amounts of a set of values for each of the plurality of the samples of the feature amounts, and a feature amount selection unit that selects a feature amount from among the plurality of the feature amounts based on a result of the principal component analysis performed by the principal component analysis unit.