Characterization of Traffic Accidents Based on Long-Horizon Aggregated and Disaggregated Data

Abstract

For sustainable transportation systems, modeling road traffic accidents is essential in order to formulate measures to reduce their harmful impacts on society. This study investigated the outcomes of using different datasets in traffic accident models with a low number of variables that can be easily manipulated by practitioners. Long-horizon aggregated and disaggregated road traffic accident datasets on Egyptian roads (for five years) were used to compare the model’s fit for different data groups. This study analyzed the results of k-means data clustering and classified the data into groups to compare the fit of the base model (Smeed’s model and different types of regression models). The results emphasized that the aggregated data used had less efficiency compared with the disaggregated data. It was found that the classification of the disaggregated dataset into reasonable groups improved the model’s fit. These findings may help in the better utilization of the available road traffic accident data for determining the best-fitting model that can assist decision-makers to choose suitable road traffic accident prevention measures.