Research Of Freeway Networks Traffic Status Identification Based On Spark

In recent years,with the scale of freeway networks expanding increasingly,traffic demand is also growing.Relying solely on freeway infrastructure construction can not effectively solve the problem of traffic imbalance between supply and demand,thus traffic congestion is serious.In this paper,aiming at inefficient single computer learning knowledge from large-scale traffic flow data,it constructs a novel model to identify the traffic status of freeway networks.The model based on Spark big data machine learning platform realize multi-object integrated traffic status identification of freeway networks.It is of great significance to improve the traffic management level of freeway networks,improve the efficiency of freeway networks operation,and provide fast,safe,smooth and intelligent travel environment.Firstly,this paper analyzes the big data processing technology and the parallel machine learning algorithm,and proposes big data machine learning platform based on Spark from the aspect of data storage,data processing and data application.Secondly,this paper utilizes the k-means algorithm to determining the relative metrics of traffic status by clustering the parameters of the freeway networks traffic flow.Since single-parameter absolute metrics of traditional traffic status identification is scientific,the proposed method selects the traffic flow,speed and occupancy as status identification parameters of freeway traffic flow.After clustering,the traffic flow data is taken as input and random forest is employed to build traffic status classification model of freeway networks.Simultaneously,in consideration of the traffic information demand of different traffic participants and freeway networks structure,the traffic index is used to quantify the traffic of the freeway networks.Finally,this paper builds a Spark big data machine learning platform,which takes the Auckland regional freeway networks as the experimental networks and uses the PeMS to collect the experimental data.It analyzes and executes the identification process of traffic status,using strategy of data parallelization and task parallelization.The results show that the parallel clustering and serial clustering have the same reliability,and the clustering results can effectively reflect the running characteristics of freeway networks traffic flow.The parallel classification and serial classification have the same accuracy.The average values of F measure,accuracy and recall rate of the random forest classification model are 98.97%,98.99% and 98.96%,respectively.In contrast to serial performance,the parallel performance of freeway networks traffic is significantly improved and the platform has desirable scalability and acceleration ratio.At the same time,the proposed traffic index can be effectively applied to the actual management of road network.