基于时空特性分析和数据融合的交通流预测

Traffic Flow Forecasting Based on Spatio- temporal Characteristic Analysis and Data Fusion

短时交通流预测是城市道路交通控制和交通诱导的关键技术之一,针对其考虑因素单一、预测精度不高的问题,提出了一种基于时空特性分析和数据融合的预测方法。首先,分析了交通流时间特性、时间相关性和基于时间序列数据的预测方法。其次,在对交通流空间特性、空间互相关性分析的基础上,提出了以相邻路段流量为自变量,采用多元逐步线性回归对目标路段流量估计预测的方法。最后,分析了交通流的时空关联特性,同时考虑到时间和空间因素,利用最小二乘动态加权融合算法将基于时间序列数据预测结果和空间回归估计预测结果进行融合输出最终结果。仿真结果表明,对比单一时间序列和空间回归估计预测方法,所提出的方法有效提高了短时交通流预测精度。

Short-term traffic flow forecasting is one of the key technologies for urban road traffic control and guidance.The prediction precision of existing methods is not high due to single consideration.A new forecasting method based on spatiotemporal characteristic analysis and data fusion was proposed.Firstly, temporal characteristics, correlation and forecasting method based on time series data of traffic flow were analyzed.Secondly, after analyzing spatial characteristic and mutual-correlation of traffic flow, a prediction method was proposed using multivariate step linear regression to estimate the target road traffic flow, which of-fered by adjacent road traffic flow as independent variable.Finally, spatio-temporal correlation characteristics of traffic flow were analyzed, considering temporal and spatial factors simultaneously.The the final result was obtained by least squares and dynamic weighted data fusion algorithm.The result of temporal was fused with spatial predictions.Compared with the former two methods, instance simulation results show that the proposed method improves the forecasting accuracy of short-term traffic flow effectively.