Optimizing Deep Ensemble Learning for Accurate Melanoma Skin Cancer Classification: Design and Analysis

Abstract

This study evaluates the performance of state-of-The-Art convolutional neural networks (CNNs) for melanoma skin cancer classification, highlighting the selection and optimization of models for ensemble learning. Wide-ResNet101-2 and resnext101-32x8d were identified as the most effective individual models based on their superior diagnostic performance metrics such as accuracy, precision, recall, and F1-score. Leveraging a weighted averaging ensemble approach, the study demonstrates a significant improvement in classification accuracy, achieving an overall accuracy of 96.12%. This advanced ensemble model surpasses traditional single-model approaches, showcasing the potential of integrated architectures in enhancing the precision of medical diagnoses. The results underscore the efficacy of ensemble learning in medical imaging, providing a robust tool for improving the detection and classification of melanoma, thereby aiding in early diagnosis and treatment.