Unveiling the potential of white shark optimization algorithm for high dimensional cancer data classification

Nature-inspired algorithms (NIA) imitate natural processes or events in order to address optimization and complicated issues. A branch of these methods known as swarm intelligence algorithms use the principle of collective behaviour of natural groups such as ant colonies, schools of fish, or flocks of birds to tackle challenging optimization issues. These algorithms use the interaction and collaboration between individual agents to efficiently explore and exploit the search space. Inspired from successful application of NIA, this research introduces White Shark Optimization (WSO) for cancer data analysis. WSO is a bio-inspired swarm intelligence algorithm that finds the best features by striking a balance between exploration and exploitation. This reduces redundancy, increases accuracy, and speeds up convergence with fewer features. This study examines the potential of WSO as a wrapper-based feature selection method in conjunction with classifiers. This analysis is conducted over four high-dimensional microarray datasets. Prior to wrapper-based classification for cancer diagnosis, filter techniques are employed as prefiltering techniques. The model uses five filter techniques, three wrappers and three classifiers. Through this combination of filter-wrapper-classifiers, 45 models are constructed. Based on the characteristics of classifiers and datasets, the model's performance changes. Thus, the optimal model is then determined using statistical analysis i.e., two-stage grading technique in which first and second stage grading results are calculated. Further, Friedman test and Wilcoxon signed-rank est is applied to validate the findings. Ultimately, the top three models are assessed. The conclusion highlights the need for an efficient model to handle high dimensional medical data.