Hybrid feature selection based on SLI and genetic algorithm for microarray datasets.

Area of Science:

• Bioinformatics
• Machine Learning
• Computational Biology

Background:

• High-dimensional microarray datasets pose challenges for machine learning due to the curse of dimensionality.
• Effective feature selection is crucial for pattern recognition, classification, and improving model performance.

Purpose of the Study:

• To present a novel and efficient hybrid feature selection method, Ga_rank&rand, for high-dimensional biological data.
• To evaluate the performance and robustness of the proposed method on standard and microarray datasets.

Main Methods:

• Developed Ga_rank&rand, a hybrid approach combining a genetic algorithm (GA) wrapper with a filter method (SLI-γ).
• Utilized SLI-γ to identify relevant features for initial solutions within the GA framework.
• Employed random features for the remaining solutions in the GA process.

Main Results:

• The SLI-γ filter method demonstrated accuracy on eleven high-dimensional datasets.
• Ga_rank&rand achieved robust performance and highly accurate solutions early in the GA evolutionary process on four microarray datasets.
• The method significantly reduced feature set size and execution time compared to a standard GA.

Conclusions:

• Ga_rank&rand is a competitive and efficient hybrid feature selection technique for high-dimensional microarray data.
• The method effectively addresses the curse of dimensionality, enhancing machine learning applications in genomics.
• Early convergence and reduced computational cost make Ga_rank&rand suitable for practical bioinformatics analysis.