A new feature selection algorithm based on relevance, redundancy and complementarity.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Accurate analysis of biological data is crucial for disease diagnosis, drug efficacy, and personalized medicine.
• Feature selection is a key technique for identifying important information from high-dimensional biological datasets.
• Existing methods often focus on relevance and redundancy, potentially overlooking synergistic feature interactions.

Purpose of the Study:

• To introduce a novel feature selection algorithm, FS-RRC (Feature Selection based on Relevance, Redundancy, and Complementarity).
• To enhance the identification of informative feature subsets by incorporating feature complementarity.
• To improve the performance of biological data analysis for various applications.

Main Methods:

• Developed FS-RRC algorithm that evaluates feature relevance to class labels, redundancy among features, and feature complementarity.
• Implemented a strategy to retain features with the highest complementarity to the selected subset.
• Compared FS-RRC against eleven established feature selection methods on synthetic and public biological datasets.

Main Results:

• FS-RRC demonstrated superior performance across key metrics including accuracy, sensitivity, and specificity.
• The algorithm exhibited enhanced stability and improved time complexity compared to existing methods.
• Experimental results confirmed the effectiveness of integrating feature complementarity into the selection process.

Conclusions:

• Integrating feature relevance, redundancy, and complementarity leads to more powerful feature subsets for biological data analysis.
• Feature complementarity plays a vital role in accurately studying biomedical phenomena.
• FS-RRC offers a robust and efficient approach for feature selection in bioinformatics.