Comparative Analysis of Feature Ranking Algorithms for Erythemato-Squamous Disease Classification

Erythemato-Squamous Diseases (ESD) are a group of skin disorders characterized by erythema (redness) and scaling. These conditions include, among others, psoriasis, seborrheic dermatitis, lichen planus, pityriasis rosea, Chronic Dermatitis, and Pityriasis rubra pilaris. Diverse classification systems have been proposed in the past to aid in the identification and management of ESDs. Traditional classification systems divide ESD into two primary categories: erythematous and squamous. They are able to distinguish between non-infectious and infectious ESDs based on the presence or absence of a fungal infection but are unable to categorize the specific type of ESD. This paper presents machine learning (ML) algorithms and state-of-the-art feature ranking algorithms for analyzing clinical data, such as patient symptoms, medical history, and visual representations of skin lesions, in order to aid in the classification and diagnosis of ESDs. The study uses state-of-the-art feature ranking algorithms such as minimum redundancy maximum relevance (MRMR), analysis of variance (ANOVA), and Kruskal-Wallis to determine the most effective features for classifying ESDs. The efficacy of ML classifiers is evaluated using a various set of best features that are identified by feature ranking algorithms. Additionally, the performance analysis is also done with different training rates. According to the analysis, Support Vector Machines (SVM) and Ensemble classifiers obtain better performance than other ML classifiers.