Skin Cancer Prediction using Deep Learning Techniques

There is a growing need for early diagnosis of skin cancer because of the rapid growth rate of melanoma skin cancer, its high treatment costs and high mortality rate. The detection of skin cancer cells was usually done manually, and most cases require a lengthy cure. Currently the main problem in skin cancer detection is high misclassification rate and low accuracy. This paper provides a technique based on deep learning techniques to detect the cancer from skin images. Convolutional neural network-based model consisting of six layers with hidden layers is used in this work. The problem of low accuracy is addressed with the help of regularization technique and features are selected with the help of convolution method. To improve the accuracy of the model hyper parameter tuning along with model parameter tuning are performed. Publicly available dataset is used in the research which contains images with cancer and normal instances. The major steps in this work includes data collection, preprocessing, data cleaning, visualization, and model development. At the end a comparative analysis is performed with state-of-the-art techniques. The proposed model achieved good accuracy of 88% on HAM dataset as compared to state of the art techniques.