Diagnosis of Skin Diseases in Children by Ensemble Machine Learning Techniques

Skin problems in children are a significant health issue and are thought to be among the most prevalent infectious diseases in children, particularly in schools. These diseases have a detrimental impact on children’s physical and mental health and can result in psychological issues that can impair their public life. These dangers can be reduced by classifying these diseases using contemporary technologies. As epidemics expand, there is an urgent need for intelligent systems that can categorize these diseases without doctor visits. By classifying skin conditions in children using ensemble machine learning approaches, which achieve higher accuracy than individual models, this study aims to contribute to a more accurate approach than earlier work. This research used ensemble learning techniques to classify skin diseases into five types (eczema, chickenpox, warts, urticaria, and impetigo). A dataset containing 4141 images collected from different sources was used, after pre-processing, it was trained by four models of ensemble learning techniques (AdaBoost, XGBoost, Random Forest, and Stacking). The results showed higher accuracy compared to previous works, as the accuracy of the AdaBoost model reached 91%, the accuracy of the XGBoost model was 92%, and the accuracy of the random forest model was 91%. The stacking model obtained the highest accuracy compared to other models, reaching 96% when combining the ResNet50 algorithm with Logistic Regression, as the use of ensemble learning with the deep learning algorithm overcame the problem of the size of the dataset used by training ResNet50 in advance on a large dataset and then adjusting the weights on a medical dataset.