Similarity Normalized Euclidean Distance on KNN Method to Classify Image of Skin Cancer

Abstract

Research on skin cancer patients in terms of optimal classification measurements is still very lacking, so it is necessary to do research that aims to get optimal values in distance measurements with normalized Euclidean distance on the KNN method to classify images of skin cancer patients. The method which is used to classify various types of data such as numbers, images, text is the K-Nearest Neighbor (KNN) method. Basically KNN, however, accepts numeric data so that data other than numeric extract them into numeric. As in this paper, the classifying images of Skin Cancer sufferers consisting of malignant and benign images is performed by extracting data with a Gray-Level Co-occurrence matrix (GLCM) to obtain numerical data from skin cancer images. The GLCM process in this paper makes the matrix be divided into contrast, dissimilarity, homogeneity, energy, correlation and ASM. Then the process is classified where the process with KNN performs the same which usually uses the Euclidean distance compared to the normalized Euclidean distance. The classification process also produces validation applying the accuracy technique calculated by MAPE. The results in this paper testing with Euclidean distance achieved MAPE of 0.71043036% by testing with Normalized Euclidean distance achieving MAPE of 0.3151053%. This showed the similarity in image classification using KNN is more optimal by using the normalized Euclidean distance approach.