Traffic Flow Prediction And State Recognition Key Technology Research Based On Cloud Computing

With the pace of urbanization process accelerating, the contradiction between rapid growth of urban road traffic demand and the slowly supply of transport infrastructure is increasingly prominent. The traffic jam is more serious in the megacities, it has seriously affected the strategy of sustainable development of urban transportation. So only accelerate the development of intelligent traffic control system with highly intelligent and information can we fundamentally solve the current urban traffic problemsThis study focused on predicting traffic flow and traffic state recognition, traffic flow forecasting and traffic state recognition technology have important research values and significance in the development and application prospects of Intelligent Transportation Systems(ITS). The purpose of this study is to provide the scientific basis and technical support for intelligent traffic management system. The main research work are as follows:1. In order to solve massive traffic of big data on real-time prediction, we introduced Hadoop platform with combining K nearest neighbor non-parametric regression algorithm to predict traffic flow. Because of the parallelism of Map Reduce framework, it greatly reduced the time to find K nearest neighbors. It is demonstrated that the prediction time on the cluster compared with that on a single machine have been greatly reduced through experiments. And the forecasting speed based on Map Reduce framework increased with the cluster size increasing, which showed good expansibility. This method can meet the need of real-time in Traffic Control and Traffic Guidance System. At present, many studies at home and abroad have focused on the time dimension to consider the traffic flow prediction, such studies ignore the impact of the road network with space dimension on the current traffic flow, this paper proposed the state vector determination method based on road network spatio-temporal correlation, so that K nearest neighbor non-parametric regression traffic flow forecasting can meet the accuracy requirement.2. This paper proposed the traffic state recognition method based on cloud computing. After clustering algorithm paralleled by Map Reduce programming model, by the effect of Hadoop platform performing parallel tasks, we can monitor the road traffic state on real time. Also K-means clustering algorithm and fuzzy C mean clustering algorithm have been improved in this paper, resulting in an initial cluster centers by Canopy algorithm, it is an effective solution for solving the blindness shortcomings of K-means and FCM randomly generate initial clustering centers. Comparative analysis of the two kinds of improved clustering algorithm used in the transport state recognition, then we selected the identification method which has higher accuracy rate as the best choice.