Procesos puntuales espacio temporales con aplicación a la modelización de Crimen : Análisis Temporal y Espacial de Robos en la Ciudad de Valencia-España.

Abstract

Valencia, the third most populous city on Spain, base its economy on industrial and agricultural activities. According to the Ministerio del Interior, the robberies is the crime with the highest incidence between 2010 and 2019. Therefore, for the purpose for analyzing whether there is any relationship between the different locations of the crimes, Moran's I test is performed. This test corroborates the assumption and allows the adjustment of a spatial econometric model of the form y=Xβ+γE, where X will be the covariate matrix and E the adjacency matrix. The methodology applied in the analysis of the security situation in Valencia is the ESF, which allows analyzing the spatial dependence of the variables in the modeling, correcting residual exogeneity errors due to spatial autocorrelation. This consists of taking a subset of eigenvectors from the adjacency matrix and adding them to the linear regression model as additional predictor variables. The selection of the eigenvectors, on the other hand, is done with the help of the Lasso method and the estimation of the fit parameter of the method with the Moran's I-test statistic which corresponds to the Cherodian contribution of the present year. The model then made it possible to determine the count of robberies recorded in each of the zones into which the city of Valencia was divided. Thus, the study identified areas that are very dangerous due to the high number of thefts recorded, even showing the influence they may have in the future in nearby areas that do not yet have this problem. The inclusion of distance covariates also made it possible to determine that the places preferred by criminals to carry out these crimes are near ATMs, banks, bars, etc.