Prediction and Classification for Melanoma Using Hybrid Learning Technique and Multi-Features

Early prediction and identification provide better chance of recovery and also reduce the mortality rate in melanoma. The initial research in the healthcare system required effective diagnosis and screening for better prediction and early analysis of skin cancer. Different methods were introduced for the prediction of melanoma. The previous methods used a single prediction stage, so the accuracy is unreliable. So, this method used a hybrid method for the prediction of melanoma. The proposed method comprises feature selection, initial prediction, and advanced prediction. The feature extraction used a gray-level cooccurrence matrix (GLCM) and threshold. The initial prediction uses the CNN method with hyperparameters, and advanced prediction is performed using an ensemble-based CNN algorithm for better prediction. The proposed work used PAD-UFES-20 dataset for implementation, and using this dataset hybrid method, it achieved 97.4% accuracy.