Pairwise Feature Selection for Multiclass Classification Beyond 10 Classes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-class classification methods struggle to robustly and accurately classify samples with multiple feature amounts into a plurality of classes, particularly in fields like biotechnology, where the number of classes exceeds 10, and the accuracy drops due to issues such as undetermined sample sizes and unclear discrimination reasons.
Innovation Solution
A method involving feature amount selection that quantifies discrimination possibilities through pairwise coupling, optimizes feature amount groups, and configures a multi-class discriminator using binary-class classifiers and a tournament hierarchy method to enhance classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing multi-class classification methods are used, then classification can be performed, but accuracy drops when the number of classes exceeds 10
Solution Approach 1:
The multi-class classification problem is segmented into multiple binary-class classification problems through pairwise coupling. Each binary classifier handles a specific pair of classes, and the results are integrated through a tournament hierarchy method. This segmentation allows the system to maintain high accuracy even when the total number of classes exceeds 10, as each binary classifier operates on a simplified two-class problem where it can achieve higher confidence.
2Reliability
If feature selection methods like t-test or information gain are used, then feature selection can be performed, but the discrimination ability for multi-class problems is insufficient
Solution Approach 1:
The feature selection process is segmented to focus on pairwise discriminatory power rather than attempting to evaluate all multi-class relationships simultaneously. By selecting features that effectively discriminate between pairs of classes and then integrating these selections through the tournament hierarchy, the method achieves reliable multi-class discrimination without the computational complexity of evaluating all possible class combinations.
Solution Approach 2:
The feature selection evaluates local discriminatory quality for each pair of classes independently. Instead of requiring features to discriminate all classes simultaneously (global quality), the method identifies features that are locally optimal for specific class pairs, then combines these locally optimized features through the tournament hierarchy to achieve global multi-class discrimination reliability.
3Adaptability or versatility
If binary-class discrimination is repeated for multi-class classification, then multi-class classification can be achieved, but it is difficult to decide comparison order
Solution Approach 1:
The tournament hierarchy method performs preliminary organization of class comparisons by establishing a hierarchical structure before actual classification. This preliminary action pre-defines the comparison order and integration path for binary classifier results, eliminating the need to dynamically decide comparison orders during operation. The hierarchy is constructed in advance based on class relationships, making the subsequent classification process straightforward and easy to execute.
4Loss of energy
If a small number of feature amounts are selected, then measurement costs are reduced, but classification accuracy may deteriorate
Solution Approach 1:
The method extracts and selects only the most discriminative feature amounts for each pairwise class comparison, rather than using all available features. By taking out only the essential features that provide maximum discriminatory power for specific class pairs and combining them through the tournament hierarchy, the system reduces measurement costs while maintaining classification accuracy. The feature selection is optimized to include only what is necessary for effective discrimination.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention is to provide a multi-class classification method, a multi-class classification program, and a multi-class classification device which select a feature amount and classify a sample into any of a plurality of classes based on a value of the selected feature amount, and a feature amount selection method, a feature amount selection device, and a feature amount set which are used for such multi-class classification. The present invention handles a multi-class classification problem involving feature amount selection. The feature amount selection is a method of literally selecting in advance a feature amount needed for each subsequent processing (particularly, the multi-class classification in the present invention) from among a large number of feature amounts included in a sample. The multi-class classification is a discrimination problem that decides which of a plurality of classes a given unknown sample belongs to.