Feature selection algorithm based on optimized genetic algorithm and the application in high-dimensional data...

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning

Background:

• High-dimensional data presents challenges in feature selection, impacting model efficiency and interpretability.
• Effective feature selection is vital for reducing dimensionality and mitigating noise interference.

Purpose of the Study:

• To propose an optimized genetic algorithm for efficient and accurate feature selection in high-dimensional data.
• To enhance model performance and interpretability through intelligent feature subset identification.

Main Methods:

• Utilized an optimized genetic algorithm simulating natural selection to search for optimal feature subsets.
• Implemented adaptive bias filtering and Gaussian code transformations for feature reduction.
• Evaluated performance across various datasets, including colon, small round blue cell tumor (SRBCT), lymphoma, central nervous system (CNS), and ovaries.

Main Results:

• Adaptive bias filtering reduced features from 724 to 372, improving accuracy from 0.9352 to 0.9815.
• Gaussian code transformation reduced features from 714 to 406, increasing accuracy from 0.9625 to 0.9754.
• Colon dataset demonstrated the highest average accuracy, with the green curve indicating stable performance within 0-300 time range.

Conclusions:

• The proposed feature selection method significantly improves the efficiency and accuracy of high-dimensional data processing.
• This approach offers a practical and effective new method for handling complex, high-dimensional datasets.
• Optimized genetic algorithms provide a robust framework for feature selection in various biological and data science applications.