Evolutionary Multitask Optimization for Multiform Feature Selection in Classification.

Area of Science:

• Machine Learning
• Artificial Intelligence
• Evolutionary Computation

Background:

• Feature selection (FS) is complex in high-dimensional spaces.
• Evolutionary computation (EC) often uses expensive wrapper-form optimization for FS.
• Filter-form optimization offers lower computational cost but focuses on feature relevance and redundancy.

Purpose of the Study:

• To propose a novel multiform optimization approach for feature selection.
• To develop an evolutionary multitask FS (EMTFS) algorithm for simultaneous optimization.
• To enhance FS efficiency and effectiveness through knowledge transfer.

Main Methods:

• Modeling FS as both a wrapper-form (accuracy-based) and filter-form (relevance/redundancy-based) optimization task.
• Developing an evolutionary multitask FS (EMTFS) algorithm for parallel task execution.
• Implementing a two-channel knowledge transfer strategy between the tasks.

Main Results:

• EMTFS achieved superior classification accuracy compared to state-of-the-art FS algorithms.
• The algorithm successfully minimized the number of selected features.
• Experimental results on public datasets validate the effectiveness of the proposed method.

Conclusions:

• Simultaneous multiform optimization is a viable strategy for feature selection.
• EMTFS effectively balances classification accuracy and computational cost.
• The proposed knowledge transfer mechanism enhances FS performance in high-dimensional data.