<![CDATA[Skin cancer is one of the most commonly diagnosed types of cancer globally. Early detection is crucial for improving the chances of recovery and preventing further complications. This study implements and compares three deep learning models—Convolutional Neural Network (CNN), Deep Neural Network (DNN), and EfficientNet—to detect skin cancer using the HAM10000 dataset. The research process includes preprocessing, model training, performance evaluation, and integration into an interactive web application based on Flask. Evaluation was conducted using accuracy, precision, recall, F1-score, and AUC metrics. The test results show that EfficientNet provides the best performance with a test accuracy of 78.44%, followed by CNN at 69.76%, while DNN only reaches 40.52% due to loss of spatial information. To improve interpretability, the system is also equipped with Grad-CAM visualization that highlights important areas in the lesion image that influence the model's decision. This study demonstrates that the EfficientNet architecture can provide more accurate and stable classification of skin lesions compared to the other two models. The practical implications of these results are the potential use of EfficientNet in clinical decision support systems to assist in the early detection of skin cancer in an automated, efficient, and accurate manner, particularly in healthcare facilities with limited resources. ]]>
Copyrights © 2025
