Abstract: Skin Disease prediction has become important in a variety of applications such as health insurance, tailored health communication and public health. Due to the costs for dermatologists to monitor every patient, there is a need for an computerized system to evaluate a patient‘s risk of melanoma using images of their skin lesions captured using a standard digital camera. The traditional diagnosis technique aims at improving the quality of existing diagnostic systems by proposing advanced feature extraction and classification methods. In the Proposed method, 40 digital images collected from AOCD unit database and another 40 digital images from MIT unit database. These images are subjected to pre-processing the images using Gaussian Filter technique. Then these images are undergone image segmentation using K-means clustering algorithm to partitioning the disease affected area and non-affected area. Feature extraction is performed using Grey Level Co-occurrence Matrix (GLCM) for examining texture which gave the statistical parameters, for better classification efficiency. The multi-SVM (Support Vector Machine) classifier is supervised learning models with associated algorithms that analyze database images for classification analysis. The diagnosis system involves two stages of process such as training and testing, the Features values of the training data set are compared to the testing data set of each type. Finally the performance analysis compared three algorithms such as Multi-SVM classifier, K-NN and Naïve Bayesian classifier. The overall accuracy of using Multi-SVM classifier is 97% to 98%.

Keywords: Skin disease prediction, GLCM, Gaussian Filter, Multi SVM, K-NN, Naïve Bayesian, K-means.