A Comparison Between SCA and ESCA Algorithms in Diagnosis an Erythema to Squamous Disease

Abstract

Dermatology is considered one of the most challenging medical specialties studied in medical schools due to the considerable similarity among various skin diseases, such as psoriasis, seborrheic dermatitis, chronic dermatitis, lichen planus, pityriasis rubra, and pityriasis rosea. With the rapid advancement of technology, computers have become deeply integrated into medicine, and many decision-support systems have been developed to assist physicians in making accurate diagnoses. Medical data obtained from laboratory analyses can play a decisive role in determining disease type. The application of classification algorithms and feature selection techniques has significantly improved the efficiency of diagnostic systems, particularly through the use of metaheuristic algorithms. In this research, a classification methodology for skin diseases is proposed by introducing a novel hybrid feature selection (FS) technique. The sine cosine algorithm (SCA) was employed within a wrapper model framework to select the optimal subset of features for classification. To enhance exploration and maintain diversity, a mutation factor was incorporated as an internal function, evolving the SCA into the enhanced sine cosine algorithm (ESCA). Consequently, the system generates two outputs for each algorithm. Experimental results demonstrated that the SCA achieved a diagnostic accuracy of 96% with 79% of selected features, whereas the ESCA achieved a remarkable diagnostic accuracy of 98% while reducing the selected features to 63%.