The Traffic Flow Prediction Based On Combination Model

The urban transport system is burdened with the task of transferring flows of people and materials in urban areas. It is the blood of a city, so its operation efficiency directly affects the entire city. In recent years, along with the social and economic development and continuous deepening of urbanization, the population and the quantity of vehicles in each city have increased constantly. Although continuous construction and perfection have been made of the infrastructure of urban transport systems, there is still a long way to go to satisfy demands of people for transport resources in urban areas. Owing to this, traffic congestion and pollution have been caused, which has brought about great inconveniences to enterprises and urban residents in their daily production and lives, thus resulting in great wastes of social resources.The ITS(Intelligent Transport System) is a comprehensive transport management system used in the 21 st century. Through monitoring, control and route guidance of operation statuses of vehicles on road, this system can optimize the reasonable distribution of traffic flows in the urban road network, improve the utilization efficiency of the urban road network, and enhance the overall transport efficiency of a city. Transport route guidance and control constitute two important parts of ITS. However, before the transport route guidance and control are made, the urban road network of a city should be forecast reasonably. In this paper, a combined traffic flow- forecasting model is designed to improve the forecasting accuracy of traffic flows.On the basis of analyzing features of urban traffic flows, a combined forecasting model for linear, non-linear and sudden traffic flows is put forward. Basic ideas of this model are as follows. A single model is used for forecasting data about traffic flows. The ARIMA model reflects the linearity of traffic flow data; the neural network is used to show the non-linearity of traffic flow data; the K-nearest neighbor nonparametric regression model is applied to reflect the suddenness of traffic flow data. On this basis, the neural network is used for forecasting and fitting of prediction results of single models.On the basis of studies made above, the combine model is used to forecast traffic flows at and in different time intervals, so that the model can provide accurate forecasting to satisfy different forecasting demands, thus improving its applicability.The IBM SPSS Modeler is used for modeling. Analyses and forecasting are made of experimental data provided by the Traffic Data Research Laboratory of the University of Minnesota-Duluth is analyzed and forecast. It is concluded that compared with common single models, the combined model can produce more accurate data.