An Effective Hybrid Model for Skin Cancer Detection Using Transfer Learning

Skin cancer is considered one of the most fatal illnesses in the human population. Within the current healthcare system, the procedure of identifying skin cancer is time-consuming and poses a potential risk to human life if not recognized promptly. Early identification of skin cancer is imperative to maximize the likelihood of achieving complete recovery. This research proposed using a hybrid model including DenseNet201 and auto-encoder for feature extraction, and a support vector machine (SVM) as the classifier. The proposed model was evaluated in the ISIC 2016 dataset, which consists of nine different classes. The hybrid model achieved a classification accuracy of 91.09% percent in accurately identifying nine different forms of skin cancer. The findings indicate that the proposed model is competitive compared to the baseline models. This outcome offers substantial assistance to dermatologists and health professionals in the field of skin cancer detection.