Skin cancer is a fatal condition. Skin cancers that because melanoma have a high fatality rate. Skin cancer can be treated and the death rate reduced for those who are diagnosed early. Since various skin lesions have many similarities, detecting skin cancer can be difficult. This study suggests a computer-based deep learning method for precisely identifying various skin lesions. Since the models in deep learning techniques study every pixel in a picture, they are particularly accurate at detecting skin cancer. The similarity of the skin lesions can occasionally confuse people, but we can reduce this by using a machine. Not all deep learning techniques, nevertheless, can produce improved predictions. Some deep learning models contain flaws that cause them to produce false-positive results. In order to categorize skin lesions and separate skin cancer from other types of skin lesions, we have implemented a number of deep learning models. The skin lesions are first classified using data preprocessing and techniques of data augmentation. Finally, to classify seven categories of skin lesions and to perform a comparison analysis, the publicly accessible benchmark ISIC 2019 dataset is subjected to the application of a salp swarm algorithm (SSA) model. By distinguishing between harmful and non-cancerous cells, the models can identify skin cancer. Performance criteria for the models include accuracy, recall, f1 score, precision, and precision recall. Finally, among all stacking models, it achieves the accuracy with the greatest value of 99%.