CCFS: A Confidence-Based Cost-Effective Feature Selection Scheme for Healthcare Data Classification.

Area of Science:

• Machine Learning
• Bioinformatics
• Data Science

Background:

• Feature selection is crucial for machine learning, particularly in healthcare data classification, but faces challenges due to large search spaces.
• Binary Particle Swarm Optimization (BPSO) is an effective evolutionary computation technique frequently applied to feature selection.

Purpose of the Study:

• To propose a novel Confidence-based and Cost-effective Feature Selection (CCFS) method using BPSO to enhance healthcare data classification.
• To improve the effectiveness and efficiency of feature selection by incorporating feature confidence and cost.

Main Methods:

• Developed a new updating mechanism within BPSO to calculate feature confidence based on feature-category correlation and historical selection frequency.
• Integrated classification performance, feature cost, and feature reduction ratio into the fitness function for comprehensive evaluation.
• Validated the CCFS method on various UCI public datasets.

Main Results:

• The proposed CCFS method demonstrated significant effectiveness in improving classification accuracy.
• CCFS effectively reduced feature selection costs while maintaining high performance.
• Experimental results confirmed the superiority of CCFS compared to benchmark feature selection schemes.

Conclusions:

• The Confidence-based and Cost-effective Feature Selection (CCFS) method offers a robust approach for healthcare data classification.
• CCFS provides an efficient and accurate feature selection strategy by balancing classification performance with feature acquisition costs.