شناسایی نقاط سیاه از طریق شبکه های بیزی کمی شاخص خطر تصادف
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|29199||2013||16 صفحه PDF||سفارش دهید||10023 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Transportation Research Part C: Emerging Technologies, Volume 28, March 2013, Pages 28–43
Traffic accidents constitute a major problem worldwide. One of the principal causes of traffic accidents is adverse driving behavior that is inherently influenced by traffic conditions and infrastructure among other parameters. Probabilistic models for the assessment of road accidents risk usually employs machine learning using historical data of accident records. The main drawback of these approaches is limited coverage of traffic data. This study illustrates a prototype approach that escapes from this problem, and highlights the need to enhance historical accident records with traffic information for improved road safety analysis. Traffic conditions estimation is achieved through Dynamic Traffic Assignment (DTA) simulation that utilizes temporal aspects of a transportation system. Accident risk quantification is achieved through a Bayesian Networks (BNs) model learned from the method’s enriched accidents dataset. The study illustrates the integration of BN with the DTA-based simulator, Visual Interactive Systems for Transport Algorithms (VISTAs), for the assessment of accident risk index (ARI), used to identify accident black spots on road networks.
Road accident statistics in Europe stress the need for more systematic mechanisms for accident analysis and prediction. According to the World Health Organization road accidents constitute one of the leading cause of death for people between the ages of 5–44 (Kapp, 2003; WHO, 2011). Given the current trends, accident fatalities are projected by 2020 to become the fifth leading cause of death worldwide resulting in an estimated 2.4 million deaths each year (WHO, 2011). At the same time, traffic accidents result in high economic losses due to traffic congestion which in turn leads to a wide variety of adverse consequences such as, traffic delays, supply chain interruptions, travel time unreliability, increased noise pollution, as well as deterioration of air quality. To combat these and the intrinsic accident risks, road safety has emerged as a priority alongside road safety management and forecasting practices. These however, suffer from major limitations and need improvement to effectively tackle this problem. One of the problems faced is data availability for the development of crash prediction and analysis models. This work contributes in this direction through the development of a prototype accident risk index quantification approach that overcomes the data availability problem by combining simulated data with historical data for the development of a BN accident prediction model. Inherently, road networks constitute complex dynamic and uncertain systems influenced by human, technological and environmental parameters. Therefore, one of the best ways to understand the causes of road traffic accidents is to develop models capable integrating significant factors relating to human, vehicle, socio-economic, infrastructural, and environmental properties. There are two broad categories of accident analysis methods: the qualitative and the quantitative. The former, despite its limited use, plays an important role in the process of accident analysis, modeling and forecasting. Qualitative analysis is subjective, exploratory and interpretative, while quantitative is based on the positivist philosophy and hence more widely used. Quantitative methods are classified into two principal groups: Time-series forecasting and Causality-based forecasting. The accident analysis approach proposed herein combines causality-based with a systemic technique, specifically, BN and Traffic simulation. The former is popular in the Artificial Intelligence domain and is based on the concept of Bayesian probability. BN provide a language and calculus for reasoning under uncertainty and incomplete information (Pearl, 2009). Hence, they are useful for inferring probabilities of future events, on the basis of observations or other evidence that may have a causal relationship to the event in question. Due to these characteristics BNs are becoming very popular in accident analysis. The second technique employed in the proposed approach is the use of a road traffic simulator using simulation-based DTA. The DTA produces estimates of traffic flow conditions for every 15-min time interval of a simulation. These estimates include, traffic flow and speed at link and movement level. Under well calibrated data, DTA estimates can be used as additional explanatory variables in accident prediction models. Overall, the method described herein quantifies accident risk index to predict road sections with high accident frequencies that constitute the network’s black spots. The paper is organized as follows: An overview of the literature and related work is presented first. Next, an outline of the method is shown. Following this, the theoretical underpinnings of BN are illustrated. Subsequently, the approach followed to implement VISTA and BN models is presented. Next, a description of the integration of the two techniques is provided and finally, results from this integration are presented before conclusions are drawn.