EEG feature selection method based on maximum information coefficient and quantum particle swarm.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Dimensionality reduction is crucial for improving electroencephalography (EEG) classification accuracy.
• Existing feature selection methods often face challenges in balancing dimensionality and classification performance.

Purpose of the Study:

• To propose an improved hybrid feature selection method for EEG data.
• To enhance classification accuracy and reduce dimensionality of EEG features.
• To optimize classifier parameters concurrently with feature subset selection.

Main Methods:

• A hybrid approach combining Mutual Information (MI) for initial feature reduction and Quantum Particle Swarm Optimization (QPSO) for optimal subset selection.
• Development of a novel fitness function considering both dimensionality and classification accuracy.
• Simultaneous optimization of classifier parameters and feature subsets.

Main Results:

• The proposed method achieved low dimensionality and high classification accuracy on EEG and UCI datasets.
• Demonstrated reduced computational complexity compared to five existing feature selection methods.
• Validated the effectiveness of the hybrid feature selection strategy.

Conclusions:

• The developed hybrid feature selection method offers a robust solution for EEG data analysis.
• The approach effectively balances feature subset dimensionality and classification performance.
• This method holds potential for various applications requiring efficient and accurate EEG signal classification.