Skin Cancer (SC) is one of the harsh diseases in humans and untreated SC will lead to death. The clinical level screening of the SC is performed using the digital-dermoscopy (DD) and evaluation of these images helps to identify the benign/malignant class infection. This work aims to propose a methodology with Shannon’s entropy (SE) pre-processing and deep-learning-based classification to detect the SC with better accuracy. The phases of the developed tool includes; data collection and resizing, executing the SE based thresholding using Mayfly Algorithm (SE + MA) to improve the visibility of the SC in DD, feature extraction using EfficientNet scheme, and classification using a chosen classifier and verifying the merit of the proposed technique based on the achieved result. In this work, 2000 images (1000 benign and 1000 malignant) are considered for the examination, and the proposed scheme helps to provide a detection accuracy of > 99% with the Random Forest (RF) classifier. This confirms the merit of the developed scheme.