Binary modified cat and mouse-based optimizer for medical feature selection: a COVID-19 case study

The cat and mouse-based optimizer (CMBO) is a metaheuristic algorithm (MA) inspired by the competitive dynamics between cats and mice. While it performs reasonably well in a range of challenges, it has limitations, such as poor exploitation capabilities and significant fluctuations in results. To address these issues, this article introduces a modified binary version of it, referred to as BMCMBO, aimed at improving performance in feature selection (FS) for medical datasets. The BMCMBO enhances the exploitation and exploration abilities of the original CMBO by integrating a pooling mechanism and three advanced search strategies: enriched chasing, fitness-based preferential selecting, and conscious escape. These modifications introduce significant changes in how search agents update their positions, how mice are selected, the role of positional information of the best agent, and the introduction of an adaptive step size. The efficiency of the BMCMBO was evaluated on 17 medical datasets. Results indicated that BMCMBO outperformed the other techniques. Additionally, the BMCMBO was applied to feature selection and diagnosis of COVID-19 in a real case study, achieving an accuracy of 98.4% in correctly classifying infected and healthy samples, further demonstrating its superior performance.