A novel feature selection algorithm based on damping oscillation theory.

Area of Science:

• Machine Learning
• Data Science
• Computational Biology

Background:

• Feature selection is crucial for reducing dimensionality in big data analysis and information retrieval.
• Existing methods often struggle to balance efficiency and accuracy in identifying optimal feature subsets.
• Noise and extraneous data can significantly degrade the performance of analytical models.

Purpose of the Study:

• To propose a novel feature subset selection algorithm, Maximum Kendall coefficient Maximum Euclidean Distance Improved Gray Wolf Optimization (MKMDIGWO).
• To enhance the efficiency of filter models and the precision of wrapper models in feature selection.
• To improve classification accuracy in big data and microarray datasets.

Main Methods:

• Developed MKMDIGWO, integrating a filter model (Kendall coefficient, Euclidean distance) and an improved wrapper model (Gray Wolf Optimization).
• Employed damping oscillation theory to dynamically adjust filter and wrapper model interactions for optimal feature subset discovery.
• Utilized support vector machine classifier for evaluating feature subset performance.

Main Results:

• MKMDIGWO demonstrated superior classification accuracy compared to four state-of-the-art algorithms across five UCI and two microarray datasets.
• The algorithm achieved at least a 0.5% higher maximum accuracy (ACC) value on 10 datasets.
• The proposed method effectively balances the efficiency of filter methods with the high precision of wrapper methods.

Conclusions:

• MKMDIGWO offers a robust and effective solution for feature selection in big data analysis.
• The integration of filter and wrapper models, guided by damping oscillation theory, leads to significant improvements in classification performance.
• This algorithm provides a valuable tool for researchers in data science and computational biology seeking to optimize feature selection processes.