Skin cancer is a common type of cancer that affects many people who have been exposed to the sun for extended periods without protection. Skin cancer is a major concern across the globe. We crafted a modified artificial neural network to sketch result for the problem. It harnesses the HAM10000 dataset for classification purposes, sorting seven skin cancer variants. A lot of models encounter difficulties at the time of classification of various types of skin cancers. Therefore, this study developed an innovative training model to reduce over-fitting allowing user interaction. This approach assists to find the optimal training duration and also avoid potential over-fitting. This proposed framework obtained accuracy of 90.62%, and the weighted average F1 score is 90.35%, respectively. Thus, our proposed framework would be able to provide a faster and efficient automatic detection of cancer.