Hybrid Convolutional Neural Network and Support Vector Machine Based on Model Optimization for Skin Cancer Classification

Skin cancer is common cancer in humans and can cause death. Handling early symptoms with the proper treatment can help patients to have a greater chance of life. The accuracy of early diagnosis in skin cancer patients plays an important role, so it needs a technique to classify skin cancer types accurately. Various deep-learning techniques for skin cancer diagnosis still require optimization to improve detection accuracy. This research proposes a hybrid Convolutional Neural Network (CNN) and Support Vector Machine (SVM) technique using pre-trained Visual Geometry Group (VGG-16) and Residual Network (ResNet-50) models and optimizing SVM parameters using grid search and a tree-based pipeline optimization tool (TPOT). The model was trained and tested using a dataset of 2000 dermoscopy images of skin cancer. This study used six experimental scenarios; each pre-trained model was trained and tested using a hybrid model with SVM. Three optimization scenarios were applied: grid search, default TPOT, and SVM-TPOT. The proposed model is proven to provide a more optimal accuracy of 91.5% on using the pre-trained model VGG-16 with the SVM-TPOT optimization method.