Feature Selection for Microarray Data Classification Using Hybrid Information Gain and a Modified Binary Krill Herd...

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Microarray datasets often contain irrelevant or redundant data, hindering accurate pattern identification.
• High dimensionality in these datasets presents a significant challenge for existing feature selection (FS) algorithms.
• Effective FS is crucial for identifying relevant attributes and improving model performance.

Purpose of the Study:

• To propose a novel and effective feature selection strategy for high-dimensional microarray data.
• To enhance the efficiency and accuracy of feature selection algorithms.
• To address the limitations of current FS methods in handling complex biological datasets.

Main Methods:

• Developed the IG-MBKH strategy, integrating Information Gain (IG) for pre-screening and an improved Binary Krill Herd (MBKH) algorithm.
• Enhanced the MBKH algorithm with a hyperbolic tangent function, adaptive transfer factor, and chaos memory weight factor for improved subset searching.
• Evaluated the strategy's performance against existing algorithms like BKH and MBKH, and assessed the impact of different classifiers.

Main Results:

• The IG-MBKH algorithm demonstrated significant improvements in convergence speed.
• Achieved a reduction in the number of selected features while maintaining or improving classification accuracy.
• Outperformed BKH, MBKH, and several recent algorithms in feature selection tasks for microarray data.

Conclusions:

• The proposed IG-MBKH strategy offers a superior approach to feature selection for high-dimensional microarray data.
• The integration of IG and enhanced MBKH effectively tackles data complexity and improves predictive model performance.
• The strategy shows promise for various classification tasks in bioinformatics and computational biology.