Epistasis-Based Feature Selection Algorithm

Area of Science:

• Computer Science
• Bioinformatics
• Chemometrics

Background:

• Variable selection is crucial for optimizing large datasets.
• Assessing gene interdependence is key for identifying significant variables.
• Epistasis concepts offer a novel approach to variable interdependence.

Purpose of the Study:

• To describe the Epistasis-based Feature Selection Algorithm (EbFSA).
• To demonstrate the efficacy of EbFSA in multivariate calibration problems.
• To compare EbFSA's performance against traditional variable selection techniques.

Main Methods:

• The Epistasis-based Feature Selection Algorithm (EbFSA) was developed.
• EbFSA was applied to multivariate calibration problems.
• Multiple linear regression models were utilized.

Main Results:

• EbFSA selected a minimal set of relevant variables.
• The algorithm demonstrated superior predictive ability compared to traditional methods.
• EbFSA achieved better model performance in multivariate calibration.

Conclusions:

• EbFSA is a powerful tool for variable selection in optimization.
• The algorithm offers significant advantages over conventional techniques.
• EbFSA enhances model predictive ability by selecting optimal variables.