路径行程时间的组合预测方法研究

A Combined Method for Prediction of Path Travel Time

立足于历史和实时数据的融合应用,从实际应用角度出发,构筑了一种路径短时行程时间的组合预测模型和相应算法。该组合预测模型包含基于历史数据特征向量的聚类分析子模型和基于路径行程时间序列的自回归子模型,通过贝叶斯概率公式实现子模型的权重分配。并对数据进行滚动式处理,实现权重系数的实时更新。最后选择上海市快速路系统3条典型路径进行实例分析,并与实际牌照自动识别行程时间数据进行对比验证。

With the implementation of the Advanced Travelers Information Systems（ATIS）,path planning has more important referenced values to the travelers and one of its core problems is estimation and prediction of dynamic path travel time.Focusing on the fusion applications of historical and real-time data,from the practical point of view,this paper aims to develop a combined short-term travel time prediction model and the corresponding algorithm.The combined method includes applications of link travel time series,and cluster analysis model which is based on feature vector and autoregressive model.In order to achieve the weight distribution and update the weight coefficient real-timely,the Bayesian approach and rolling horizon processing on the data are utilized.Finally,three typical paths of Shanghai expressway system are selected as the theoretic test cases,and the forecast results are validated with the field automatic plate recognition travel time data.