Mining Of Traffic Flow Based On Decision Tree Algorithm And Clustering Analysis

Urban road traffic is an important part of intelligent transportation system, its biggest feature is that intersection is complex and the traffic flow of intersection influences each other. There is no doubt that the running state of intersection will be directly decided to the degree of congestion.The research for intersection has become the focus in entire road network.Based on the huge amounts of traffic flow data which are accumulated by the intersection of urban road,using the data mining technology to analyze it and find the implicit model of traffic flow,providing technical support for easing traffic pressure,optimizing transportation network and implementing intelligent management of road.Firstly,this paper analyses the basic characteristics of urban road's traffic flow. On the basis of investigating and analyzing traffic flow of intersection, using hierarchical clustering and K-means clustering algorithm respectively to cluster and analyse data of traffic flow that has been collected to make the traffic flow of sequence which has strong correlation between each other get together for the same class and obtaining spatial distribution feature of urban roads;Secondly,this paper compares clustering effect of two clustering methods and improves hierarchical clustering algorithm to make the improved clustering algorithm more scalable. Based on the results of clustering analysis of the traffic flow data,this paper preprocesses and joins the relevant properties of traffic flow to generate corresponding training set, taking improved C4.5 algorithm of decision tree to classify the training set and generate road congestion classifier. The classifier improves the classification of the timeliness and the accuracy of the prediction. Finally the generated classification is used in real-time dynamic traffic flow data to forecast traffic congestion of intersection.The experimental results prove that the improved clustering algorithm and decision tree algorithm which are used to analyse traffic running state of intersection is feasible and can get good accuracy of prediction.