关联交叉口交通流模糊变权重组合预测方法

针对单一交通流预测方法存在的局限性和传统交通流组合预测模型中权重不能动态变化的问题,提出一种关联交叉口交通流模糊变权重组合预测方法。先对交叉口交通流的关联性进行分析,并给出关联交叉口的定义;再建立关联交叉口交通流模糊自适应变权重组合预测模型,该模型分别利用Kalman滤波器模型与SVM模型来预测关联交叉口交通流量,然后根据这2个模型预测的误差和交通量的变化趋势,采用模糊逻辑推理方法,对这2个预测模型分别赋予适当的权重。试验结果表明,组合预测模型的最大绝对误差、平均绝对误差和相关系数均明显好于单一的预测方法,分别为9.8%、4.63%和0.99。

A Fuzzy Variable Weight Combination Algorithm for Forecasting Traffic Flow of Related Intersections

Seeing that limitations there were in various single traffic flow forecasting methods and weights in the traditional combination forecasting model can’t dynamically change,a fuzzy variable weight combination forecasting algorithm for traffic flow of related intersections was proposed.First,the related intersection was defined by analyzing the relationship between traffic flow.Then,a nonlinear combined forecasting model based on method of fuzzy adaptive variable weight was built.It uses the Kalman filter and SVM to forecast the traffic flow successively and gets the appropriate weights of each model with fuzzy logic.The results show that the maximum absolute error of the combined forecasting model is 9.8%,the average absolute error is 4.63% and the correlation coefficient is 0.99,which are significantly better than that of single model.