A Multi-Scenario Identification Strategy for High-Risk Vehicles Based on Vehicle Networking Data

Accurate identification of high-risk vehicles is a prerequisite for effectively mitigating traffic risks. Based on the vehicle networking data, including integrated information on vehicle attributes, daily operational patterns, and driver behavior indicators from multiple systems such as speed and mileage monitoring, GPS, ADAS and DSM, this study proposes a multi-scenario identification strategy for high-risk vehicles by dividing related risks into mileage risks and driving behavior risks. First, based on the proven independence of the mileage risk indicators, according to the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$3\sigma $</tex-math> </inline-formula> principle, the numeric characteristics of high-risk mileage is excavated, and the single-risk scenario and mixed-risk scenario are designed to evaluate the mileage risks, and the results show that single mileage risks are more common and typical than mixed risks. Second, as for driving behavior risks, an improved Hidden Markov Model is used to process sparse driving behavior warning data, and combined with association rule mining and zero-inflated Poisson regression, a behavior risk chain is constructed to trace the formation of severe danger states. Results reveal that vehicle spacing and distraction are key triggers of high-risk states. Compared with a single comprehensive evaluation method, this method can highlight the process of formation process of driving risk formation and is more targeted in achieving risk control.