The Application of the K-Medoid Classification Method for Analyzing Crime Rates in South Sulawesi

Abstract

This study employs the k-medoid clustering method to analyze districts and cities in South Sulawesi based on their crime rates. As the population increases, employment opportunities may decline, potentially elevating stress levels and, consequently, the likelihood of criminal behavior. To evaluate the distribution of criminal incidents across South Sulawesi, the k-medoid method is used to classify regions into clusters. Unlike other clustering methods, k-medoid utilizes the median as the cluster center (medoid), making it more robust to outliers. Specifically, the Partitioning Around Medoids (PAM) algorithm is applied, in which initial objects are randomly selected to represent clusters. If the error value is high, the cluster centers are iteratively adjusted until the error is minimized. The dataset consists of crime incidence data for South Sulawesi in 2020, encompassing various types of crimes. Based on the Silhouette coefficient, the optimal number of clusters was determined to be three: Cluster 1 comprises 11 regions, Cluster 2 includes 8 regions, and Cluster 3 contains 5 regions. These clusters provide a comprehensive overview of the crime patterns across different regions within the province.