An Evaluation Performance of Kernel on Support Vector Machine to Classify The Skin Tumors in Dermoscopy Image

Skin cancer has recently become one of the types of cancer that often appears and could become deadly. Mortality from skin cancer patient could be reduced if the detection and treatment is early and appropriate. Segmentation of skin lesions is usually on images that have classified melanocytic, whereas skin lesions that are classified as nonmelanocytic are equally important. Support vector machine (SVM) are used to differentiate skin lesions in dermoscopic images. The results of the classification, achieving best performance with accuracy of 85%, sensitivity of 86%, specification of 84%, and precision of 88% using radial basis function kernel. RBF kernel is giving best performance for this type of data. For validation model, this study using k-Fold Cross Validation. The optimal value are k=7 and k=8 with an accuracy of 83%. This study gives an idea to deal with disease which related to skin cancer using image processing technique.