Application of Spatial Econometrics Analysis for Traffic Accident Prediction Models in Urban Areas

The previous researches on the prediction of accidents frequently assumed the independence among the error terms, from the standpoint of traditional statistics. However, as spatial data including information about geographical spaces, data collected in each traffic analysis zones are not randomly distributed in space and have the spatial correlation each other. That violates the basic assumption. To control such autocorrelation, spatial econometrics analyses need to be considered. The aim of this study is to identify the spatial correlation of traffic accidents and to develop prediction models with spatial econometrics analysis in urban areas. As a result, it was found that traffic accidents revealed high spatial correlation and spatial econometrics models showed a more enhanced explanatory power than normal linear regression model. Moreover, spatial econometrics models were more excellent than it, when verifying it through SRMSE.