A novel nonlinear time-varying sigmoid transfer function in binary whale optimization algorithm for descriptors...

Area of Science:

• Computational Chemistry
• Bioinformatics
• Machine Learning

Background:

• High-dimensional molecular descriptors cause the curse of dimensionality in computational chemistry.
• Feature selection is crucial for improving model performance and interpretability.
• Binary whale optimization algorithm (BWOA) is a promising metaheuristic for feature selection.

Purpose of the Study:

• To introduce a new version of the nonlinear time-varying Sigmoid transfer function to enhance the standard whale optimization algorithm (WOA).
• To propose a novel BWOA algorithm, BWOA-3, specifically designed for molecular descriptor selection.
• To evaluate the performance of BWOA-3 in solving descriptor selection problems and improving drug classification.

Main Methods:

• Development of a new nonlinear time-varying Sigmoid transfer function for WOA.
• Introduction of the Binary Whale Optimization Algorithm version 3 (BWOA-3).
• Validation using a high-dimensional drug dataset and comparison with other binary optimization algorithms (BWOA variants, BBA, BGWOA, BMRFO).

Main Results:

• BWOA-3 demonstrated superior performance in descriptor selection compared to other algorithms.
• The proposed algorithm achieved faster convergence and selected smaller feature subsets.
• Significant improvements were observed in classification performance metrics, including accuracy, specificity, sensitivity, and f-measure for Amphetamine-type Stimulants (ATS) drug classification.

Conclusions:

• BWOA-3 is highly effective for molecular descriptor selection, mitigating the curse of dimensionality.
• The enhanced Sigmoid transfer function improves exploration and exploitation in WOA.
• BWOA-3 offers a superior approach for enhancing drug classification accuracy in computational chemistry.