Predictive accident modeling for highway transportation system using Bayesian networks

Chen, Dan (2014) Predictive accident modeling for highway transportation system using Bayesian networks. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (2222Kb)

Abstract

The highway network, as a critical infrastructure in our daily life, is an important component of the public transportation system. In the face of a continuously increasing highway accident rate, highway safety is certainly one of the greatest concerns for transportation departments worldwide. To better improve the current situation, several studies have been carried out on preventing the occurrence of highway accidents or reducing the severity level of highway accidents. The principal causes of highway accidents can be summarized into four categories: external environment conditions, operational environment conditions, driver conditions and vehicle conditions. This research proposes a representational Bayesian Networks (BNs) model which can predict and continuously update the likelihood of highway accidents, by considering a set of well-defined variables belonging to these principal causes, also named risk factors, which directly or indirectly contribute to the frequency and severity of highway accidents. This accident predictive BNs model is developed using accidents data from Transport Canada's National Collision Database (NCDB) during the period of 1999 to 2010. Model testing is provided with a case study of Highway #63 site, which is from 6 km southwest of Radway to 16 km north of Fort Mackay in north Alberta, Canada. The validity of this BNs model is established by comparing prediction results with relevant historical records. The positive outcome of this exercise presents great potential of the proposed model to real life applications. Furthermore, this predictive BNs accident model can be integrated with a Safety Instrumented System (SIS). This integration would assist in predicting the real-time probability of accident and would also help activating risk management actions in a timely fashion. This research also simulates 10 scenarios with different specific states of variables to predict the probability of fatal accident occurrence, which demonstrates how the BNs model is integrated with SIS. The major objective of this research is to introduce the predictive accident BNs model with the capabilities of inferring the dependent causal relations and predicting the probability of highway accidents. It is also believed that this BNs model would help developing efficient and effective transportation risk management strategies.

Item Type: Thesis (Masters)
URI: http://research.library.mun.ca/id/eprint/8080
Item ID: 8080
Additional Information: Includes bibliographical references (pages 56-64).
Keywords: Bayesian network (BNs), highway safety, predictive accident model, Safety Instrumented System (SIS)
Department(s): Engineering and Applied Science, Faculty of
Date: October 2014
Date Type: Submission
Library of Congress Subject Heading: Traffic safety--Mathematical models; Traffic accidents--Forecasting--Mathematical models; Bayesian statistical decision theory--Data processing; Neural networks (Computer science)

Actions (login required)

View Item View Item

Downloads

Downloads per month over the past year

View more statistics