Skin cancer is one of the dangerous cancers that threatens human life and can be seen in lesions on the patient’s skin. If this disease is not treated in time, it can threaten the patient’s life. However, with early detection of this disease, more effective treatment measures can be used and its progression can be prevented. Many machine learning and deep learning methods have been developed for early detection of skin cancer. But one of the advantages that can be provided by deep learning is its ability to learn from large data sets, which makes it more favorable for deep learning methods. The present paper proposes a method based on the combination of Swin transformer and Wavelet for classifying skin cancer images. In addition, feature optimization is performed using improved gray wolf optimizer(IGWO), Fox optimizer (FOX) and modified gorilla troops optimizer (MGTO) to identify optimal feature values. ISIC-2016 and ISIC-2017 datasets are used to evaluate the developed model. From the obtained results, it is shown that the Swin Transformer-based Wavelet +MGTO model provides the best result on the ISIC-2016 data set, achieving an accuracy of 0.9792. Furthermore, the Swin Transformer-based Wavelet +IGWO model provides the highest performance on the ISIC-2017 dataset, with an accuracy of 0.9792 and an F-measure score of 0.9780 on the ISIC-2017 data set.