Batik classification using KNN algorithm and GLCM features extraction

Informatics Department, Sanata Dharma University, Yogyakarta, Indonesia

^* Corresponding author: rita_widiarti@usd.ac.id

Abstract

Batik is one of the Indonesian cultures that has been recognized by UNESCO as an intellectual right of Indonesia. The popularity of batik internationally raises concerns about the Indonesian people’s understanding of batik if Indonesian people only refer to all types of batik just as ‘batik’. By utilizing K-Nearest Neighbour (KNN) algorithm which is a simple classification algorithm, then a system can be created that can classify batik types. The first step of KNN is training, which stores each training pattern. The second step is classification, whenever classifying a pattern, KNN examine all training patterns to determine the K closest patterns using certain calculations such as Euclidean Distance and Manhattan Distance. Before classification, a characteristic that represents a pattern is needed. Gray-Level Co-Occurrence Matrix (GLCM) is an algorithm that has proven to be very powerful as a feature descriptor in representing the texture characteristic of an image. This research experiments with the value of K in KNN = 1, 3, 5, and 7 with the distance calculation using Euclidean and Manhattan. The GLCM characteristic used are Entropy, Energy, Contrast, Homogeneity, Dissimilarity, Correlation, ASM, and the average of each characteristic. From the research that has been done, the system created obtained the highest accuracy of 75% with the combination of parameters; pixel distance = 7, K value = 1, 1st fold as test data and 2nd and 3rd fold as training data, and by using StandardScaler. But despite getting decent accuracy, there is still a need for further research to improve the accuracy of batik classification.