Performance of majority voting transfer learning deep learning monkeypox disease detection

Abstract

Global health concerns have been raised by the advent of monkeypox following the COVID-19 epidemic, highlighting the need for reliable automated systems to support early skin disease screening. This study proposes a monkeypox skin disease classification framework using a majority voting ensemble based on transfer learning. The ensemble combines predictions from multiple pretrained convolutional neural networks (CNNs) to improve classification robustness. A publicly accessible dataset comprising four classes: monkeypox, chickenpox, measles, and normal skin was used for the experiments. The results indicate that while a single ResNet50 model achieved the highest overall accuracy (99.15%), the majority voting approach yielded higher precision than several individual models, demonstrating improved reliability in distinguishing monkeypox cases. These findings suggest that ensemble-based majority voting can enhance the robustness of monkeypox skin disease classification and may support computer-aided screening systems.