Clustering Modeling of Divorce-Causing Factors Using the K-Means Algorithm

Abstract

The increasing divorce rate in North Sumatra is a complex social problem. Traditional policy approaches often generalize root causes across regions, especially by assuming economic issues as the main trigger. Unlike previous studies that group regions solely by geographical proximity or raw case volume, this research introduces a novel approach that clusters regions based on interaction patterns among substantive causal variables. This research applies the K-Means clustering algorithm to map the dominant factors contributing to divorce across 33 districts/cities in North Sumatra, using 2023-2025 Statistics Indonesia (BPS) data. The clustering model was evaluated using the Elbow Method and Davies-Bouldin Index (DBI). The optimal evaluation resulted in 3 clusters () with a highly valid DBI score of 0.62. The clustering revealed three regional typologies: extreme multi-crisis, high-conflict, and low-case regions. The results showed that ongoing disputes consistently dominate as the primary cause across all clusters, significantly surpassing economic factors. Practically, these clustering results provide a critical, data-driven foundation for policymaking. Rather than providing generalized financial assistance, local governments and Religious Courts can use this model to deploy targeted, region-specific psychological interventions and revitalize Pre-Marriage Guidance (Bimwin) programs tailored to each region's crisis profile.