Integrated Use of Rough Sets and Artificial Neural Network for Skin Cancer Disease Classification

The aim of this study is to build a classifier for predicting a disease existence by learning a least conventional set of features extracted from the clinical dataset. Rough set, indiscernibility relation method along with a feedforward neural network is applied and divided the whole process into two parts. At first part, the rough set method is considered as a reduction of features and selected as subset of attributes. In the next part, classification via feedforward artificial neural network is applied to the selected reduction on the dataset. Obtaining datasets of skin cancer disease from the Engineering in Medicine and Biology Society (EMBC) has been used to test the classifier. Our proposed method obtained 95% accuracy for melanoma skin cancer detection. In this regard, this (ANN) model is proposed intended for detecting automatically the cancer patients at a primary stage. Finally, our proposed model is working improved as opposed to some other conventional model (for example RF and SVM).