A Deep Convolutional Neural Network (DCNN) with Fine Tuned Hyper Parameters for Skin Cancer Classification

Skin cancer is increasingly becoming a prominent community health hazard across the globe, and detecting its onset very early gives an edge in advancing patient care quality. This work introduces a novel approach, an advanced convolutional neural network model rooted in deep learning to effectively classify skin lesions, with the primary focus on improving accuracy by fine tuning the hyper parameters. The model is fitted and assessed using the HAM10000 dataset. The dataset has 10,015 dermoscopic pictures, encompassing a range of skin manifestations. To enhance the accuracy of our model, we employed several techniques, including batch normalization, dropout, data augmentation, and data balancing. The proposed model outperformed several existing models and achieved an impressive accuracy of 96% in classifying skin lesions, demonstrating its role as a key asset in assisting dermatologists and clinicians in diagnosing skin cancer.