Explainable AI framework for skin lesion imaging based cancer detection

The classification of skin cancer remains a challenging task because of the subtle visual patterns of benign and malignant lesions. In this work we propose an interpretable ensemble learning framework for dermoscopic skin lesion classification by combining Random Forest, XGBoost and LightGBM using Max-Voting. The proposed system demonstrates good predictive performance with clinical interpretability, achieving 95.94% accuracy on the HAM10000 dataset. To increase transparency we use explainable AI techniques to visualise the image regions and features that are most important for the model's decisions. We also perform feature selection based on Genetic Algorithm to select the most discriminative descriptors and to reduce the redundancy in the hybrid feature space. The combination of deep features, handcrafted features, ensemble learning and explainability leads to a strong framework for accurate prediction and meaningful interpretation. The experimental results show that the proposed approach outperforms the individual classifiers and provides more explainable clinically relevant explanations that may help to improve the confidence and trust of dermatologists. In summary, the study demonstrates that the combination of ensemble learning and explainable AI can provide an effective and practical direction towards trustworthy skin cancer decision support.