Optimizing Skin Cancer Prediction: An Efficient Deep Learning Approach

Skin cancer is a prevalent and possibly fatal disease that needs to be appropriately diagnosed and treated in its early stages. The International Skin Imaging Collaboration (ISIC) provided training and testing images of benign and malignant skin lesions. We have employed a multi-layer convolutional neural network to classify skin cancer and compared the results with other models, such as random forest, support vector machine, and ResNet. Data augmentation and rescaling are two data preparation methods employed to improve the model’s performance and generalization. Batch normalization and dropout regularization were used to prevent over-fitting and stabilize the training process. On the held-out test set, the proposed model achieved an accuracy of \(94.9\%,\) surpassing the accuracy of other models like ResNet \((90.51\%)\) , random forest \((90.51\%),\) and support vector machine \((85.3\%).\) The proposed system aims to provide dermatologists with an automated tool to assist in early skin cancer detection and classification to improve patient outcomes and speed up the diagnosis process with high accuracy.