The development of cutting-edge, scalable, and trustworthy approaches in particular has been the emphasis in order to address problems like the precise diagnosis of malignant melanoma in radiographs. Effective treatment and prognosis for melanoma depend on prompt identification. To fulfill the needs of modern healthcare, there is a worldwide physician shortage, which causes data imbalance problems across several healthcare sectors. Due to these imbalances, deep learning algorithms frequently give particular data groupings. As we know that Machine learning doesn’t identify unbalanced classes. Our work suggests a ground-breaking deep learning detection method for skin cancer employing an unbalanced dataset by aligning several malignancy categories. Different subtypes of melanomas make up the dataset, called Skin Cancer MNIST: Ham10000, and deep learning methods are often used for disease categorization using imaging. The findings demonstrate that on the unbalanced dataset, CNN with ADASYN outperforms VGG-16, VGG-19, and parallel CNN in views of accuracy, F1-score, and Reca11. The suggested method's accuracy, Precision, Reca11, and F1-score values were 93.84%, 92.77%, 90.54%, and 91.67%, respectively. By comparison, the other techniques' accuracy values were 79.45%, 83.04%, 69.57%, and 71.19%, respectively. Our suggested strategy could aid in identifying diseases, which might prevent deaths, lessen the need for unneeded biopsies, and lower expenditures for patients, dermato1ogists, and hea1th care workers