Multi-Class Skin Lesion Classification Using EfficientNet-B7 and Ensemble Learning

Melanoma skin cancer is a major global health problem, and this makes early diagnosis an essential element in treatment. In this research work, a multi class skin lesion classification model based on EfficientNet-B7 along with ensemble approaches is developed. The objective is to improve the predictive power for skin lesion detection systems, while incorporating the best aspects of deep learning techniques as well as ensemble learning. In order to extract a set of representative features, EfficientNet-B7 is applied for the CNN model as the feature extraction technique, and ensemble learning is a powerful technique which combines multiple models to attain higher accuracy, prevent overfitting, and heave more numbers of classes. The advantage of the proposed methodology is illustrated by the results of experiments with a large skin lesion database showing better diagnostic accuracy as compared to individual models and current literature approaches. The performance of the proposed ensemble model for lesion type classification is summarized in accuracy, precision, and recall, and F1-score with an overall of 94.8%. It is shown that the ensemble technique improves the melanoma, basal cell carcinoma, squamous cell carcinoma, and benign classification remarkably. The conclusion of this research will utilize potential solution for the automated detection and diagnosis of skin lesions this paper thus offer a reliable equipment to clinicians, in preventing and managing skin lesions through early detection. Possible subsequent work may include enhancingletalizing onto more modalities for improved performance of the system and their relevance for clinical practice.