Modelling the Amount of Crime in Indonesia Using Geographically Weighted Negative Binomial Regression Approach

Abstract

Acts of criminality is the act of someone who may be liable to punishment on the basis of the criminal code or
the laws and other regulations in force in Indonesia. So far the increase and decrease of crime tend to be small, but the
averAGE NUMBER OF CRIMINAL ACTS IN Indonesia is still very high. Various theories concerning the cause of the
crime have been proposed by experts from various disciplines and fields of science, but so far it is still not well there is
one answer to a satisfactory settlement. According to the theory of cartography/geography, crime spread into a region
geographically or socially. It means there is a spatial aspects in the spread amount of crime through social relations or
human interaction. Modeling problems of criminality in Indonesia cannot be done globally because according to the
theory of spatial elements of cartography/geography is very noteworthy. Regression analysis involving the geographical
aspect known as spatial regression. Method of spatial regression approach that is often used to model data in the form of
response variables divide tubers (count) is a geographically weighted poisson regression (GWPR). In the application
model GWPR often data do not meet the assumption that average and variansi are the same, but often variansi is greater
than the mean, the case is usually called overdispersi. Negative binomial regression (NBR) model is one of the models
that could be used to resolve the case of the overdispersi. NBR model involving the location called geographically
weighted negative binomial regression (GWNBR). In this paper we discussed the amount of crime in Indonesia modeling
using GWNBR approaches. The data used are secondary data sourced from publications BPS 2014 that includes 31
provinces in Indonesia. The results obtained show that the GWNBR model can cope with case overdispersi. The results
of a test of suitability of the model shows that the model GWNBR in accordance. There are five predictors variables, i.e.
the percentage of the poor population (ð‘¿ðŸ), the open unemployment rate (ð‘¿ðŸ), percentage of population aged 10 years or
over who have never school (ð‘¿ðŸ‘ ), overcrowding (ð‘¿ðŸ’), and per-capita gross regional domestic product ( ð‘¿ðŸ“ ) jointly
significant effect against the amount of crime in Indonesia(ð’š). Test result predictor variables individually shows that the
variables ð‘¿ðŸ , ð‘¿ðŸ‘ , ð‘¿ðŸ’ , and ð‘¿ðŸ“ each significantly influential in all provinces in indonesia, while the predictor variables ð‘¿ðŸ
do not affect significantly in the province of NAD and province of North Sumatra.
Keywords: Crime, Overdispersion, GWPR, NBR, GWNBR.