Deep Learning-Based Skin Lesion Classification with Ensemble Stacking and Data Augmentation

The classification of skin lesions using deep learning models has made significant advancements, particularly in enhancing the early detection and diagnosis of skin cancer, such as melanoma. However, the challenges posed by class imbalance and variability in medical image datasets limit the generalization capabilities of traditional models. In this paper, we propose an approach that integrates ensemble stacking with data augmentation techniques. By combining the strengths of multiple pretrained models, this architecture improves classification accuracy and robustness against heterogeneous data. The proposed method was evaluated on the ISIC 2018 dataset, achieving high accuracy and demonstrating improved performance compared to individual models. Our approach offers a reliable solution to enhance skin lesion classification, with potential applications in broader medical imaging tasks.