Analysis of Traffic Accident Risk in Nganjuk Using Point Process on a Linear Network with Lasso Regularization

The Nganjuk Region of East Java has a very high incidence of traffic accidents, largely attributable to the high traffic density. To conduct a comprehensive analysis of the risk of traffic accidents in the Nganjuk Region, in this study we employ detailed data involving the coordinate locations of traffic accidents integrated with environmental factors, road geometrical features, and remote sensing. To perform the analysis, we consider the Poisson point process model on a linear network. Furthermore, we combine the standard maximum likelihood technique with Lasso regularization to perform variable selection and parameter estimation simultaneously. We develop the algorithm by considering three techniques consecutively namely the Berman-Turner approach, Taylor expansion, and coordinate descent algorithm (CDA). Among the six covariates used in this study, we select three that are significant for the intensity of risk accidents in the N ganj uk Region: distance to the nearest traffic light location, road types, and number of lanes. This information is useful for the Nganjuk Region government as a consideration in formulating policies to prevent traffic accidents and to provide useful information to the community and the government regarding locations prone to traffic accidents.