Research On The Technology Of Urban Road Traffic State Identification Based On Hadoop

With the rapid development of social economy, the urban road congestion problems have become increasingly prominent, which has brought great distress to the city residents’ travel life. As an important part of intelligent traffic management system, traffic state identification technology accurately identifies the road traffic state, to provide reliable information to the induced traffic participants, so as to effectively alleviate the traffic congestion problem. At present, the road network of China’s large and medium-sized city is very large, which produces a large number of traffic data all the time. Traditional single machine has been unable to deal with it effectively. Therefore, this paper uses the Hadoop platform which has the Strong parallel computing power to excavate the useful information from the massive traffic data to improve the accuracy of traffic state identification.Taking the intelligent traffic management and control platform project of Nansha Free Trade Zone in Guangzhou as research background, this paper makes a deep study of the urban road traffic state identification technique based on Hadoop, including the following aspects:1) This paper systematically introduces the traffic big data, and describes the acquisition and pre-processing technology of the traffic data in detail. Besides, we describe the data mining technology and analyze the related technologies of Hadoop platform in detail.2) We study deeply how to select the characteristic indexes of the traffic state,and select speed,flow and occupy rate as the input data of the traffic state identification algorithm.Since the FCM algorithm randomly select the initial clustering center and does not fully consider the different contribution of the data attributes, we use the K-means algorithm and the feature weighting method to improve the FCM algorithm.After completing the parallelization design of the improved FCM algorithm, we use the improved FCM algorithm to analyze the historical traffic data on the Hadoop platform. At last, the improved FCM algorithm are evaluated and compared from the error rate and the speedup ratio. The experimental results show that the improved FCM algorithm is the best.3) According to the historical traffic data which is identified by the improved FCM algorithm, we use the random forest algorithm to predict and classify the traffic state. To improve the operational efficiency of the random forest algorithm, the establishment process and prediction process of the random forest algorithm is parallel designed. Besides, we use the random forest algorithm to predict and classify the traffic state on the Hadoop platform,and analyze the speedup ratio of the random forest algorithm.At the same time, the random forest algorithm, the Bayes algorithm and the SVM algorithm are evaluated from the the accuracy rate on the Hadoop platform. The results show that the accuracy of the random forest algorithm for traffic state identification is the highest, reaching 91.1%. Finally, the running time of the random forest algorithm on single machine and Hadoop platform is analyzed and compared.4) We introduce the traffic cloud platform in detail from the framework design and the function realization. Besides, we design the establishment process of the traffic state identification system based on Hadoop platform. At last, we achieve the practical application of the traffic state identification technology based on Hadoop platform in the display of the realtime traffic state, the forecast of the future traffic state, the query of the historical traffic state and the analysis of the traffic state.