Skin cancer can be a very serious and widespread disease that is affecting many individuals of all ages and can be caused by excessive contact with UV rays or even artificial sources. Early diagnosis is crucial in treating skin cancer effectively. However, traditional methods of diagnosis, such as biopsy, can be painful, slow, and time-consuming. In recent years, the line of computer-based skin cancer detection has been revolutionized by the application of deep-learning algorithms and models. In this paper, we have worked on Skin Cancer MNIST: HAM10000 data. The study presented here shows that it is possible to classify skin cancer using deep-learning algorithms. The findings demonstrate that deep-learning models, particularly MobileNet, can reliably identify skin cancer and can also classify the skin lesions. The models outperformed different deep-learning algorithms, in terms of accuracy and precision when their performance was compared. The goal is to determine how well deep learning does in relation to highly trained dermatologists. In general, the means suggest that all deep-learning models performed better than dermatologists. Our MobileNet model is outstanding among the other fifteen models as it has an accuracy of 72% compared to other GRU, LSTM, and CNN models. In future work, we can work on imbalanced datasets.