Citation

Ramírez AF, Valencia C. Accid. Anal. Prev. 2020; 149: e105848.

Copyright

(Copyright © 2020, Elsevier Publishing)

DOI

10.1016/j.aap.2020.105848

PMID

33166761

Abstract

The planning and location of resources for urban traffic management generate complex decision problems, given the uncertainty of variables that explain traffic behavior, the lack of data, and the large number of factors to be considered when creating optimal policies. In particular, the attention given to traffic-related accidents by local authorities requires the modeling and forecasting of events that are spatial and temporally defined. In this study we use data from the traffic police department in Bogota (Colombia) about incidents with injuries or fatalities between 2014 and 2016. We locate each event in spatial coordinates and crossed the observations with exogenous variables (climate, seasonal events and road properties). We model the spatiotemporal stochastic process for accidents using a Log-Gaussian Cox model given its flexibility as it enables the use of fixed and random effects. The results of this study are summarized in three main contributions: (i) identification of principal factors that increase the risk of traffic accidents and fatalities; (ii) identification of critical zones in the city that require more attention; and (iii) a predictive tool to forecast accidents in the future, including the stochastic properties that can be used in a prescriptive model.

Language: en

Keywords

Bayesian inference; Log-Gaussian Cox process; Spatiotemporal model; Traffic accidents prediction