Skin cancer is the most serious health problems in the globe because of its high occurrence compared to other types of cancer. Melanoma and non-melanoma are the two most common kinds of skin cancer. One of the most difficult problems in medical image processing is the automatic detection of skin cancer. Skin melanoma is classified as either benign or malignant based on the results of this test. Impediment due to artifacts in dermoscopic images impacts the analytic activity and decreases the precision level. In this research work, an automatic technique including segmentation and classification is proposed. Initially, pre-processing technique called DullRazor tool is used for hair removal process and semi-supervised mean-shift algorithm is used for segmenting the affected areas of skin cancer images. Finally, these segmented images are given to a deep learning classifier called Deep forest for prediction of skin cancer. The experiments are carried out on two publicly available datasets called ISIC-2019 and HAM10000 datasets for the analysis of segmentation and classification. From the outcomes, it is clearly verified that the projected model achieved better performance than the existing deep learning techniques.
