基于模糊逻辑的山区高速公路交通事故预测研究

Research on Prediction of Traffic Accidents onMountainous Highways Based on Fuzzy Logic

为解决事故预测中变量及变量关系之间的模糊性,引入模糊逻辑方法,建立基于模糊逻辑的山区高速公路交通事故预测模型。收集云南省某典型山区高速公路的道路几何线形和交通事故等数据,按道路是否位于山阴面进行分类,提出能表征特殊节点和不良天气影响程度的显著变量;将变量划分为5个模糊集合,并采用三角形、梯形隶属函数将变量模糊化,训练得到693条有效模糊控制规则,按照模糊推理机制对模型进行推理,反模糊化后得到精确的事故预测结果。结果表明,模型的预测精度、平均绝对误差和平均绝对百分比误差分别为90.2,0.88次/km,11%,预测精度较高且适用于同类型的山区高速公路。因此,利用模糊逻辑对山区高速进行事故预测具有一定的可行性和可移植性。

To solve the fuzziness between variables and variable relationships in accident prediction,fuzzy logic is introduced to establish a fuzzy logic-based traffic accident prediction model for mountainous highways.The data of road geometric alignment and traffic accidents of a mountainous highway in Yunnan Province are collected,and the data are also classified according to whether the road is located on a shaded side of the mountain to propose significant variables that can characterize the degree of influence of special nodes and bad weather.The variables are divided into five fuzzy sets,and triangular and trapezoidal affiliation functions are used to fuzzify the variables,and the variables are trained to obtain 693 effective fuzzy control rules.Then the model is reasoned according to the fuzzy inference mechanism,and the accurate accident prediction results are obtained after anti-fuzzification.The results show that the prediction accuracy,average absolute error and average absolute percentage error of the model are 90.2%,0.88 times/km and 11%,respectively,and the prediction accuracy is high and applicable to the same type of mountain highway.Therefore,the use of fuzzy logic to predict accidents on mountainous highways has a certain degree of feasibility and portability.