The Short-term Traffic Flow Prediction Of Main Thoroughfares Of City

With the development of socio-economic and the acceleration of urbanization, the number of motor vehicle keeps sharp increasing, resulting in not only the traffic congestion, traffic accidents, but also causing serious environmental pollution. In this context, Intelligent Transportation Systems (ITS) come into being, it combines advanced electronic information technology, data communications technology, control technology and so on, and builds a kind of intelligent traffic control and management system, this system can play a comprehensive role in a wide range. The core subsystems of ITS include advanced traffic information service subsystem (ATIS) and advanced traffic management subsystem (ATMS). The real-time, accurate traffic flow prediction, as part of the traffic parameters, provides an important technical support for the realization of the two major subsystems. The city’s main thoroughfares, as objects, are focused on in this paper to study the short-term traffic flow prediction methods; it proposes some improvements on the basis of previous studies and gets some good results in final.After analysing the temporal characteristics of traffic flow, a kind of combination prediction model is put forward in the paper. The future projections are dynamically adjusted according to the current traffic flow data in the first part. Meanwhile, through the analysis of spatial and temporal characteristics of historical traffic flow data, the historical curve which is similar to the current traffic flow characteristics is sought in another part to find the data that is matching to the predicted value. Then, the information obtained by the both sides can be combined to achieve the final short-term traffic flow forecasting.In the foundation of above research, the non-parametric regression forecasting method is studied, too. The DTW distance is used as measurement criteria to calculate the distance, which overcomes the matching problem caused by the telescoping of the time series on the time axis.Using the traffic flow of Xiamen Lotus junction cross-section as instance, it is demonstrated by the simulation experiments of the prediction methods that the maximum absolute relative errors of the combination prediction methods are all below30%, and the mean absolute percentage errors are less than9%, which indicat the predicted effects are good; the prediction accuracy of the improved non-parametric regression method:s better than the unmodified one’s.The prediction software is realized by using the Visual Basic language and has been applied to the intelligent traffic control system of Xiamen.