Theft Prediction Model Based on Spatial Clustering to Reflect Spatial Characteristics of Adjacent Lands

Previous studies have shown that when a crime occurs, the risk of in adjacent areas increases. To reflect this, previous grid-based prediction combined all cells surrounding event location to be predicted for use model training. However, actual land is continuous rather than set independent as geographic information system. Because patterns occur according detailed method vary, it necessary spatial characteristics prediction. In this study, with similar were classified using Max-p region (a clustering technique), and performance was compared existing random forest tree-based machine learning model). According results, F1 score increased by approximately 2%. Accordingly, there are differences physical environmental factors influenced crime. The findings reveal involving same offender likely around area original crime, indicating repeated features where occurred.