A quasi-optimal channel selection method for bioelectric signal classification using a partial Kullback-Leibler...

Area of Science:

• Machine Learning
• Data Science
• Biomedical Engineering

Background:

• Effective variable selection is crucial for accurate classification in high-dimensional datasets.
• Existing methods may not efficiently identify the most informative variables for specific tasks.

Purpose of the Study:

• To propose a novel variable selection method utilizing a new metric, the partial Kullback-Leibler (KL) information measure.
• To evaluate the contribution of each variable (dimension) in the data for improved classification.

Main Methods:

• Estimated probability density functions using a multidimensional probabilistic neural network trained on KL information theory.
• Defined partial KL information measure as a ratio before and after dimension elimination.
• Selected effective dimensions by iteratively eliminating ineffective ones based on the partial KL information.

Main Results:

• Applied the method to channel selection for nine subjects, reducing channels by 54.3 ±19.1%.
• Achieved an average classification rate of 96.6 ±2.8% using only three or four selected channels.
• Demonstrated high classification performance with reduced data dimensionality.

Conclusions:

• The proposed partial KL information measure is effective for variable and channel selection.
• The method enables significant data reduction while ensuring optimal or quasi-optimal classification accuracy.
• This approach is valuable for applications requiring efficient and accurate classification from high-dimensional data.