DEEP LEARNING BASED CLASSIFICATION MODEL WITH DATA AUGMENTATION FOR SKIN CANCER DETECTION

Abstract

Skin cancer is a major global public health concern, accounting for roughly 2.1 million new cases diagnosed globally each year. Improving survival rates requires early discovery and treatment, but one major obstacle is the scarcity of dermatologists in isolated areas. The application of deep learning and artificial intelligence to the prediction of skin cancer has increased dramatically in the last few years. This work investigates the wide range of machine learning algorithms used in this context and does a thorough assessment of sophisticated skin cancer prediction methods using deep learning techniques. Because of the overlapping phenotypic features, dermatologists face a tremendous task while diagnosing skin cancer, which consists of seven separate diagnoses. The range of 62% to 80% is normal for conventional diagnostic accuracy, highlighting the potential of machine learning to improve diagnosis and treatment. Although some researchers have developed binary skin cancer classification algorithms, it has been difficult to expand this to additional classes with better results. We created a deep learning classification model for different forms of skin cancer, and the findings show that deep learning is better at classification jobs. The results of the experiment show that when the accuracy of the CNN and Sequential models is combined, the Image Processing model produces the best accuracy. With an astounding 98% accuracy rate, the Image Processing model outperforms a comprehensive dermatological examination in terms of precision. Additionally, a comparison with the most recent skin categorization models highlights how much better the suggested multi type model for classifying skin cancer.