Application Of The Dbscan Algorithm For Clustering Criminal Evidence Data At The Bengkulu District Attorney's Office

Abstract

Evidence management at the Bengkulu District Attorney's Office was previously handled manually, which complicated storage and retrieval as caseloads increased. This study implements the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to group 150 evidence records based on evidence type, quantity, date received, and storage location. A quantitative-experimental design was used, covering data pre-processing, encoding, normalization, and clustering with Python on Google Colab, using epsilon = 0.8 and minimum points = 3. DBSCAN produced four clusters and identified eight records as noise, achieving a Silhouette Score of 0.4757, indicating moderate clustering quality. The clusters were respectively dominated by electronics, vehicles, sharp weapons, and documents, each linked to a separate storage warehouse. These findings show that DBSCAN can support more structured and efficient evidence storage and retrieval governance within the prosecutor's office.