A Hybrid Neural Network Approach Using SOM and LVQ for Mapping Crime Clusters in Indonesia

Abstract

Crime ratehigh crime rates in Indonesia are one of the important issues that need to be addressed with data-based strategies. This study aims to group provinces in Indonesia based on crime patterns using Self-Organizing Map (SOM) and classify the results using Learning Vector Quantization (LVQ). The results of the clustering analysis using SOM show that the optimal number of clusters is two, as supported by validation using Connectivity, Dunn Index, and Silhouette Score. Cluster 1 consists of 31 provinces with lower crime rates, while Cluster 2 includes 3 provinces with higher crime rates. To improve understanding of the clustering results, classification was carried out using the LVQ method, which produced an accuracy of 91.43%.