SVM-RFE based feature selection and Taguchi parameters optimization for multiclass SVM classifier.

Area of Science:

• Machine Learning
• Bioinformatics
• Data Mining

Background:

• Support Vector Machine (SVM) is effective for binary classification but limited in multiclass problems.
• Feature selection and parameter optimization are crucial for improving SVM performance in complex datasets.

Purpose of the Study:

• To enhance multiclass classification accuracy using Support Vector Machine (SVM).
• To investigate the efficacy of combining SVM Recursive Feature Elimination (SVM-RFE) with the Taguchi method for optimizing SVM parameters (C and γ).

Main Methods:

• Applied SVM-RFE to select optimal feature subsets from the Dermatology and Zoo datasets.
• Utilized the Taguchi method to optimize the C and γ parameters of the SVM classifier.
• Evaluated classification accuracy on multiclass problems using selected features and optimized parameters.

Main Results:

• Feature selection via SVM-RFE identified key variables, improving classification performance.
• Taguchi parameter optimization significantly boosted SVM accuracy for multiclass classification.
• Achieved classification accuracy exceeding 95% for both the Dermatology and Zoo datasets.

Conclusions:

• The combination of SVM-RFE and Taguchi method effectively addresses SVM limitations in multiclass classification.
• This integrated approach offers a robust strategy for improving diagnostic accuracy in medical and biological datasets.
• The findings demonstrate a significant advancement in applying machine learning for complex classification tasks.