An augmented Snake Optimizer for diseases and COVID-19 diagnosis

Area of Science:

• Computational intelligence and machine learning applied to medical informatics.
• Development of novel algorithms for data analysis in healthcare.

Background:

• Feature Selection (FS) is crucial for enhancing classification performance in medical applications.
• Existing metaheuristic algorithms require optimization for effective feature selection in complex medical datasets.

Purpose of the Study:

• To introduce two novel wrapper-based FS approaches utilizing the Snake Optimizer (SO) metaheuristic.
• To enhance the binary version of SO (BSO) with evolutionary crossover operators to create BSO-CV.
• To evaluate the efficacy of BSO and BSO-CV on COVID-19 and other disease benchmark datasets.

Main Methods:

• Development of a binary Snake Optimizer (BSO) using an S-shape transform function for discrete feature values.
• Integration of three evolutionary crossover operators (one-point, two-point, uniform) into BSO to create BSO-CV.
• Comparative analysis of BSO and BSO-CV against standard BSO and recent FS methods on multiple datasets.

Main Results:

• BSO-CV significantly outperformed standard BSO in accuracy and running time across 17 datasets.
• BSO-CV achieved an 89% dimension reduction on the COVID-19 dataset, compared to BSO's 79%.
• BSO-CV demonstrated improved exploration-exploitation balance, leading to better convergence towards optimal solutions.

Conclusions:

• The enhanced BSO-CV algorithm shows significant potential for effective feature selection in medical datasets.
• BSO-CV offers a robust and efficient method for reducing data dimensionality while maintaining high classification accuracy.
• The proposed approach provides a promising tool for advancing machine learning applications in disease diagnosis and analysis.