Comparison of three classifiers used in the detection of benign tumor and malignant melanoma skin diseases

It is an unknown fact that many skin diseases have similar type of shape, size and symptoms. Hence, it is a cumbersome task to recognize and classify these diseases by the doctors. So, for the correct identification of skin disorders, doctors need to check the patient’s history alongside certain laboratory testing and physical examinations. But all these processes are time consuming and also costlier for a common man. Hence, this paper discusses a MATLAB based software system introduced to reduce the complexity and thereby providing accurate results. This system includes image preprocessing, features extraction and classification for prediction of the type of skin disorders. Besides feature extraction, the paper mainly focusses on the classification based on three classifiers—SVM (Support vector machine), KNN (K- nearest neighborhood) and NB (Naïve Bias classifier)—and provides a comparative result based on various parameters. It can be concluded from the comparison tables that among the three classifiers, SVM provides the highest accuracy of 98.73% while KNN with 93.67and and NB with 84.81%. This classification helps a doctor to achieve the exactness of the type of skin disorder. In this system the patient needs to provide the image of the infected portion as input and the proposed system shall detect the disease.