Optimal Method for Traffic Accident Case Retrieval and Evaluation

Xianyuan Dong,Shouen Fang,Junhua Wang,Shuo Liu,Pin Wang
2013-01-01
Abstract:An optimal method for traffic accident case retrieval has been proposed to improve accuracy of case retrieval. This method aims to assist users in decision-making in real-time under traffic accident conditions. And a new evaluation index called matching degree, as supplement of case similarity, is proposed for assessing the retrieved case set comprehensively. Different from previous studies, the presented optimal method and evaluation index have the following four critical features: (1) proposing an information entropy method to evaluate dispersion of traffic accident data and determining the weight value of each traffic accident case feature objectively; and (2) establishing a traffic accident case retrieval base which contains several sub-case bases, and that puts a solid foundation for equalizing case retrieval; then (3) proposing a global similarity model of traffic accident cases and two equations for calculating local similarity of numerical traffic data and categorical traffic data; and (4) presenting a new evaluation index called matching degree firstly, make up deficiency of the original evaluation index. This study develops a prototype system for traffic accident case retrieval experiments. Dozens of case retrieval experiments have demonstrated the promise of this optimal case retrieval method and evaluation method for traffic accident management in real-time.
What problem does this paper attempt to address?