Diagnosis of Skin Cancer Using Feature Engineering Techniques

Major mortality rate among human beings is due to cancer. Early diagnosis of Skin cancer especially Melanoma is having the potentiality to reduce morbidity as the major reason behind the disastrous repercussions of three out four homo-sapiens is due to skin cancer. Detection of cancer using machine learning and deep learning algorithms makes it very much feasible and economical. The ultimate focus of this paper is for detecting skin cancer at an early stage and helping to combat the increasing cases in skin cancer patients. In this paper, we have implemented different types of CNNs of different configurations on categorical classification where architectures were trained on different input image size and selecting of best architecture was based on various metric evaluations like Maximum Accuracy, Precision, Recall, and F1 score and best architecture has achieved high accuracy and performed outstandingly in all the evaluation section. Architecture 4 performed overall excellent in terms of every field of metric evaluations. This architecture will be a helpful tool for diagnosing skin cancer at an early stage and will take the less computational cost for classifying the skin cancer disease.