Citation

Lavrenz SM, Vlahogianni EI, Gkritza K, Ke Y. Accid. Anal. Prev. 2018; 117: 368-380.

Affiliation

Lyles School of Civil Engineering, Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907, United States. Electronic address: ke10@purdue.edu.

Copyright

(Copyright © 2018, Elsevier Publishing)

DOI

10.1016/j.aap.2017.11.030

PMID

29530303

Abstract

The use of statistical models for analyzing traffic safety (crash) data has been well-established. However, time series techniques have traditionally been underrepresented in the corresponding literature, due to challenges in data collection, along with a limited knowledge of proper methodology. In recent years, new types of high-resolution traffic safety data, especially in measuring driver behavior, have made time series modeling techniques an increasingly salient topic of study. Yet there remains a dearth of information to guide analysts in their use. This paper provides an overview of the state of the art in using time series models in traffic safety research, and discusses some of the fundamental techniques and considerations in classic time series modeling. It also presents ongoing and future opportunities for expanding the use of time series models, and explores newer modeling techniques, including computational intelligence models, which hold promise in effectively handling ever-larger data sets. The information contained herein is meant to guide safety researchers in understanding this broad area of transportation data analysis, and provide a framework for understanding safety trends that can influence policy-making.

Copyright © 2017 Elsevier Ltd. All rights reserved.

Language: en

Keywords

Computational intelligence models; Crash data modeling; Econometric methods; Statistical methods; Time series analysis; Traffic safety