Melanoma is well-known skin cancer that cause fatal. Therefore, detection of melanoma at early stage are essential to enhance the successful of survival rate. For the detection of melanoma, proper analysis is carried out on the skin lesion according to a set of specific clinical characteristics. This skin lesion clinically diagnosed begin with primary clinical screening and dermoscopic analysis, a biopsy and histopathological examination. Lastly, this skin lesion is classified as either "potential melanoma" or "non-melanoma”.However, detection of skin cancer in the early stages is a difficult and expensive process. Typically, the analysis to checks for the various Melanoma are using pre-defined thresh-olds in classification stage such as Asymmetry, Border, Colour, Diameter and Evolution (ABCDE) where color, texture, size and shape are being analysis for image segmentation and feature stages. Accuracy for this method was encourage and reach up to 95.45%.The proposed method DL shows best accuracy when compared with other methods. To our best knowledges, we are not aware of any previous work proposed for this task. The proposed deep learning frameworks were evaluated on the ISIC 2017 testing set.