Classification of Erythematous - Squamous Skin Diseases Through SVM Kernels and Identification of Features with 1-D Continuous Wavelet Coefficient

Abstract

Feature extraction is a kind of dimensionality reduction which refers to the differentiating features of a dataset. In this study, we have worked on ESD_Data Set (33 attributes), composed of clinical and histopathological attributes of erythematous-squamous skin diseases (ESDs) (psoriasis, seborrheic dermatitis, lichen planus, pityriasis rosea, chronic dermatitis, pityriasis rubra pilaris). It's aimed to obtain distinguishing significant attributes in ESD_Data Set for a successful classification of ESDs. We have focused on three areas: (a) By applying 1-D continuous wavelet coefficient analysis, Principle Component Analysis and Linear Discriminant Analysis to ESD_Data Set; w_ESD Data Set, p_ESD Data Set and LESD Data Set were formed. (b) By applying Support Vector Machine kernel algorithms (Linear, Quadratic, Cubic, Gaussian) to these datasets, accuracy rates were obtained. (c) w_ESD Data Set had the highest accuracy. This study seeks to identify deficiencies in literature to determine the distinguishing significant attributes in ESD_Data Set to classify ESDs.