Skin malignancy is a catastrophic health problems witnessed in Europeans and western area of the world because of the changes in the ozone layer. Ultraviolet (UV) rays common threats for the human health. Scientists have studied on Computer-Aided Diagnosis (CAD) scheme to ease interpretation detection of melanoma. There are several variations of features of the lesions and different Artificial Intelligence (AI) based design participates in an essential role for building CAD system. This study has refined skin lesions with diffusion and dull razor technique. Lesion images have taken for color-based shape and texture feature extraction. Scientists have found new fused color features are effective for melanoma and nevus classification. It has discovered details of 2000 images from ISIC (database archive) helped to build improved feature set. These features were analyzed through various 12 machine learning models as highest accuracy of 93.9%. Proposed Deep Neural Network (DNN) has reached 95.8% accuracy within few epochs. This model was assessed specific limits which were discussed in the results section. In future this exercise will motivate investigators to experience with color features and its variations with other AI based models.