Improved Sine Cosine Algorithm for Global Optimization and Medical Data Classification

Abstract

The Sine Cosine Algorithm (SCA) is a powerful metaheuristic algorithm that has been efficiently employed to solve complex optimization problems, including data classification. However, it suffers from the challenges of a poor search mechanism and a lack of effective exploration–exploitation trade-off. Thus, it is necessary to propose effective strategies to tackle the shortcomings of SCA. This study introduces the improved IMSCA using the interactive multi-leader search mechanism and search process leading in multiple directions. The performance of IMSCA in solving eighteen classic benchmark functions and thirty functions of the CEC-2017 test suites in both 10 and 100-dimensional modes was evaluated and compared to efficient algorithms and improved SCA variations. The statistical tests demonstrated that IMSCA outperformed the other algorithms in most cases. IMSCA was also used for the evolution of support vector machine (SVM) parameters and weighting the features of eight real-life medical datasets for data classification. The empirical results proved the superior performance of IMSCA, with the highest classification accuracy and an FMR of 1. Furthermore, IMSCA was utilized for COVID-19 diagnosis in a case study. It differentiated between infected and uninfected individuals with an accuracy rate of 98%. This implies that IMSCA is a promising algorithm.