GEOGRAPHICALLY WEIGHTED GENERALIZED POISSON REGRESSION AND GEOGRAPHICALLY WEIGHTED NEGATIVE BINOMIAL REGRESSION MODELING ON PROPERTY CRIME CASES IN CENTRAL JAVA

Abstract

Property crime in Indonesia remains one of the most prevalent categories of crime across various regions of the country. This category encompasses a range of criminal acts, including theft, illegal appropriation of goods, robbery, motor vehicle theft, arson, and property damage. One of the commonly used regression analysis methods is Poisson regression. The assumption violation of overdispersion in Poisson regression is often found in property crime data in Central Java. This study also considers spatial aspects, depicting local regional characteristics and the integration of local and global variables. Therefore, this study employs Geographically Weighted Generalized Poisson Regression (GWGPR) and Geographically Weighted Negative Binomial Regression (GWNBR) methods with Adaptive Bisquare Kernel weighting. The aim of this research is to develop a model for each district/city in Central Java using Adaptive Bisquare Kernel weighting, thus providing a more accurate representation of the factors influencing property crime in each region. The AIC value criterion of 411.3652 indicates that the GWNBR method is the most suitable for modeling the number of property crime cases in each district/city in Central Java compared to Poisson regression, negative binomial regression, and GWGPR methods.