Predicting crash risk and identifying crash precursors on Korean expressways using loop detector data

https://doi.org/10.1016/j.aap.2015.12.004Get rights and content

Highlights

  • We developed crash risk prediction models on Korean expressways.

  • The factors affecting on crash risk differed by segment type and traffic flow state.

  • We investigated on the effect of the detector stations at irregular intervals.

  • The significant traffic variables were selected by conditional logistic regression.

  • Separate models to predict crash risk were developed using genetic programming.

Abstract

In order to improve traffic safety on expressways, it is important to develop proactive safety management strategies with consideration for segment types and traffic flow states because crash mechanisms have some differences by each condition. The primary objective of this study is to develop real-time crash risk prediction models for different segment types and traffic flow states on expressways. The mainline of expressways is divided into basic segment and ramp vicinity, and the traffic flow states are classified into uncongested and congested conditions. Also, Korean expressways have irregular intervals between loop detector stations. Therefore, we investigated on the effect and application of the detector stations at irregular intervals for the crash risk prediction on expressways. The most significant traffic variables were selected by conditional logistic regression analysis which could control confounding factors. Based on the selected traffic variables, separate models to predict crash risk were developed using genetic programming technique. The model estimation results showed that the traffic flow characteristics leading to crashes are differed by segment type and traffic flow state. Especially, the variables related to the intervals between detector stations had a significant influence on crash risk prediction under the uncongested condition. Finally, compared with the single model for all crashes and the logistic models used in previous studies, the proposed models showed higher prediction performance. The results of this study can be applied to develop more effective proactive safety management strategies for different segment types and traffic flow states on expressways with loop detector stations at irregular intervals.

Introduction

With the tremendous growth of Intelligent Transportation System (ITS) during the past decade, it is now possible to collect traffic parameters such as traffic volume, speed, and occupancy from a variety of detectors in real time. Several studies have attempted to demonstrate the potential application of loop detector data in order to reveal the relationship between crash occurrence and traffic parameters (Hughes and Council, 1999, Oh et al., 2001, Oh et al., 2005, Lee et al., 2002, Lee et al., 2003). Also, proactive safety management strategies utilizing Advanced Traffic Management Systems (ATMS) such as variable speed limits (VSL) and ramp metering have showed effects on improving traffic safety (Abdel-Aty et al., 2006, Abdel-Aty et al., 2007, Lee et al., 2006a, Lee et al., 2006b, Lee and Abdel-Aty, 2008). Within these studies, real-time crash risk prediction models were estimated to predict crash occurrence likelihood based on real-time traffic data.

The crash precursors mean traffic flow characteristics leading to crashes, which are identified by comparing the crash cases and non-crash cases. The crash precursors also will be different depending on traffic flow characteristics by surrounding environment on expressways. Especially, these traffic flow characteristics are divided by segment type and traffic flow state in Highway Capacity Manual (HCM). Therefore the real-time crash risk prediction models based on the crash precursors are required to develop for different segment types and traffic flow states on expressways. However, the majority of studies have developed a single model (e.g., aggregated model) for overall crashes (Abdel-Aty et al., 2004, Abdel-Aty et al., 2008, Abdel-Aty and Pande, 2005, Ahmed and Abdel-Aty, 2012, Ahmed and Abdel-Aty, 2013, Yu and Abdel-Aty, 2013). A major drawback of the aggregated modeling technique is that it cannot consider different effects of traffic flow characteristics on crash risk by roadway geometry and traffic flow states. Therefore, separated crash risk prediction models were developed by segment type and traffic flow state on expressways in this study. Also Korean expressways have irregular intervals between loop detector stations. Therefore, we investigated on the effect and application of the detector stations at irregular intervals for the crash risk prediction on expressways.

Data used in this study were collected from the Gyeongbu expressway in Korea. The real-time traffic data (volume, speed and occupancy) was obtained from loop detectors and matched with historical crash data. The real-time crash risk prediction models were developed by segment type and traffic flow state. The mainline of expressways is divided into two segments based on ramp presence, basic segment and ramp vicinity, and the traffic flow states are classified into uncongested and congested conditions. Genetic programming technique was used in this study to develop the real-time crash risk prediction models. The genetic programming is a relatively new modeling technique that was proposed to solve classification and regression problems. Compared with traditional statistical regression methods and machine learning algorithms, two major advantages of genetic programming were proposed (Xu et al., 2013b). First, genetic programming can find a solution to a problem without any pre-specified functional forms. Second, in contrast with different machine learning algorithms, genetic programming can remove the “black box” effect and make the model understandable. Due to genetic programming models lack the ability to select significant variables that play a role to increase the reliability of prediction models, conditional logistic regression analysis was first estimated to select the most significant traffic variables contributing to crash occurrence. The conditional logistic regression analysis has an advantage which can control confounding factors by matching. Based on the chosen explanatory variables by conditional logistic regression analysis, the genetic programming models have been estimated and the relationship between traffic variables and crash risk has been investigated for each condition. Finally, the prediction performance of the proposed models have been compared to single model based on Receiver Operating Characteristics (ROC) curves and areas under the ROC curve (AUC).

Section snippets

Background

Real-time crash risk prediction models were estimated with the purpose of identifying the crash precursors and the results were applied in proactive safety management strategies. With the advanced traffic surveillance system (loop detectors, remote traffic microwave sensors, automatic vehicle identification systems), traffic flow characteristics prior to crash occurrence could be identified and matched with the crashes. Abdel-Aty et al. (2004) developed crash likelihood prediction model for

Data collection

To estimate the relationships between traffic flow characteristics and crash risk, crash data and traffic data were collected from the Gyeongbu expressway in Korea. The Gyeongbu expressway, connecting Seoul to Busan, is the longest and most heavily traveled expressway in Korea. The entire length from Seoul to Busan is 416.0 km, and the yearly 371 million vehicles traveled through this expressway in 2010. The Gyeongbu expressway has irregular intervals between loop detector stations. Table 1

Conditional logistic regression analysis

The matched case-control structure was used in this study. It was adopted to identify the significant traffic variables leading to crash occurrence while controlling for time of the day, day of the week, season, and location (i.e., geometric characteristics). Therefore, conditional logistic regression analysis is used in variable selection process, it is expected to provide accurate results as the effects of confounding factors are controlled by matching. The modeling is estimated under the

Preliminary analysis

A preliminary analysis of characteristics for crashes used in this study was conducted to investigate the crash mechanism by segment type and traffic flow state on expressways. Table 3 presents the results of the preliminary analysis. On the basic segment, the percentage of crashes caused by unsafe speed, lack of visual attention, and following too closely under the congested condition is larger than that under the uncongested condition. Also, the crashes associated with multiple vehicles and

Conclusion

In order to improve traffic safety on expressways, it is important to develop proactive safety management strategies, such as providing crash risk information and controlling traffic flow to reduce crash risk in real time, with consideration for segment types and traffic flow states because crash mechanisms had some differences by each condition. Hence, this study aimed to identify crash precursors which mean the traffic conditions leading to crash occurrence and to develop separate crash risk

References (33)

  • C. Lee et al.

    Evaluation of variable speed limits to improve traffic safety

    Transp. Res. Part C

    (2006)
  • A. Pande et al.

    Assessment of freeway traffic parameters leading to lane-change related collisions

    Accid. Anal. Prev.

    (2006)
  • C. Xu et al.

    Evaluation of the impacts of traffic states on crash risks on freeways

    Accid. Anal. Prev.

    (2012)
  • C. Xu et al.

    Predicting crash likelihood and severity on freeways with real-time loop detector data

    Accid. Anal. Prev.

    (2013)
  • R. Yu et al.

    Utilizing support vector machine in real-time crash risk evaluation

    Accid. Anal. Prev.

    (2013)
  • M. Abdel-Aty et al.

    Assessing safety on Dutch freeways with data from infrastructure-based intelligent transportation systems

    Transp. Res. Rec.: J. Transp. Res. Board

    (2008)
  • Cited by (0)

    View full text