Classification of Underdeveloped Areas in Indonesia Using the SVM and k-NN Algorithms

Abstract

The determination or classification of underdeveloped areas essentially consists of classifying several observations taking into account existing indicators. The classification method used is K-Nearest Neighbor (k-NN) and Support Vector Machines (SVM). This study aims to analyze the accuracy of the classification between SVM and k-NN algorithms in the classification of underdeveloped areas in Indonesia. The data source used in this study is secondary data obtained from the Central Bureau of Statistics (BPS). The data used are 514 districs and municipalities of Indonesia. After analysis, the conclusion is that there are 122 districs and municipalities that are left behind out of a total of 514 districs and municipalities in Indonesia. The most underdeveloped areas are on the island of Papua, followed by the areas of the islands of Bali and Nusa Tenggara, and Sulawesi. Based on the results of the classification of underdeveloped areas using the method SVM with the kernel RBF has the best results with the parameters C = 1 and γ = 0.05 while the results of the classification of underdeveloped areas using the method k-NN obtains the best results with k = 15 Based on the results of classification of underdeveloped areas using the SVM and the k-NN method, including the level of classification is very good. The two methods compared have the same precision value of 92.2% and can be used to determine the classification of underdeveloped areas. Keywords: classification, machine learning, supervised learning, underdeveloped areas.