Classification of Skin Disease using Machine Learning

Abstract

Erythemato-squamous disease (ESD) is one of the dermatology field's complex diseases. Due to its common morphological features, it is challenging to diagnose and generally produces inconsistent results. In addition, the physician's expertise was used to make the diagnosis based on the observed symptoms. The accurate classification of erythemato-squamous disorders is one of the dermatology field's problems that need attention, and to help with this issue, by using clinical and histopathological data, this tool will differentiate the six classes of ESD. In this research, we have applied 3 different machine learning algorithms as base models i.e. Random Forest, Decision Tree, and Naïve Bayes to classify the ESD and 5 Ensemble Meta techniques such as Voting classifier, average classifier, Stacking, boosting, and bagging classifiers to measure the accuracy. In comparison to other classifier methods, the ensemble technique employed on dermatology dataset, original dataset and clinical feature extraction to identify which model performs better on both cases. The ensemble method provides a more precise and accurate prediction of skin diseases.