Performance Comparison of Machine Learning Based Classifiers for Melanoma Cancer Diagnosis

Abstract

In recent years one among the rising deadliest diseases is skin cancer. Skin cancer is one among the foremost difficult illness among numerous cancer kinds. This paper proposes an automated skin lesion analysis system for the first detection and classification of melanoma using image process techniques. First, dermoscopy image of skin is taken as image acquisition step, then pre-processing step for noise removal and post-processing step for image improvement. Identification of the diseased/abnormal portion of the skin is possible solely by correct delineation ways. Therefore here the processed image undergoes image segmentation. Second, options are extracted using feature extraction technique ABCD parameter, GLCM, and FOS. A comparison of the performance of all feature sets is conferred during this paper so as to see what feature sets offer the most effective classification results. Numerous feature combinations are given because the input to the KNN, SVM & ANN classifiers. Performance is analyzed supported the accuracy of the learning classifier output.