SCIENTIA GEOGRAPHICA SINICA ›› 2017, Vol. 37 ›› Issue (2): 209-216.doi: 10.13249/j.cnki.sgs.2017.02.006

Special Issue: 地理大数据

• Orginal Article • Previous Articles     Next Articles

Distributed Incremental Traffic Flow Big Data Forecasting Method Based on Road Network Correlation

Xin Li1(), Deyou Meng1   

  1. 1.Collaborative Innovation Center of Three-aspect Coordination of Central Plain Economic Region, Henan University of Economics and Law, College of Resource and Environment, Henan University of Economics and Law, Zhengzhou 450046, Henan, China
  • Received:2016-02-25 Revised:2016-08-05 Online:2017-02-25 Published:2017-02-25
  • Supported by:
    National Nature Sciences Foundation of China (41501178), Henan University of Economics and Law Dr. Startup Funds(800257)


Along with the accelerating urbanization, there are more and more contradictions between the number of cars and urban transportation facilities. The congestion time and congested roads in cities are increasing. Intelligent urban traffic management platform is the effective method to alleviate the increasingly serious urban congestion problems. By using prediction results of traffic flow big data, the platform can guide users to adjust the travel plan, and ease the traffic pressure effectively. How to use a large number of spatio-temporal data related to traffic activities to predict the traffic flow is the key to realizing traffic guidance. In this article, a distributed incremental aggregation method for traffic flow data is studied. The method combines the distributed incremental data aggregation method with the traffic flow data cleaning rules, makes cleaning and counting of traffic flow big data, and provides data for traffic flow forecast. With the analysis of traffic flow correlation in the network of upstream and downstream, this article uses the multi-order of turning rate in the intersection to quantize the correlation, builds the spatial weight matrix based on the road network correlation, and improves the STARIMA model. In this article, two groups of contrast experiments were made. Through the contrast experiment between MapReduce method and MPI method, the result proves that the method proposed in this article is better than the MPI method in the development cycle and stable operation. The method’s efficiency can meet the need of traffic flow data aggregation. The traffic flow statistics can be used as the basis of traffic flow forecasting. Through the contrast experiment between the Improved STARIMA model and the Dynamic STARIMA model, the result proves that the Improved STARIMA model, which considers the multi-order correlation between the upstream and downstream sections, matches the distribution rules of traffic flow in road network better. Therefore, the forecast results are more accurate. In conclusion, the method of this article is a new method of traffic flow forecasting in the background of big data, and it can realize accurate prediction.

Key words: traffic flow, big data, distributed incremental, road network correlation, STARIMA

CLC Number: 

  • K909