Research And Application Of Traffic State Identification Algorithm In Big Data Environment

Along with the quick development of economy and society,the problem of congestion on urban roads has become serious increasingly,which has not only increased people’s travel time and the commuting costs,but also caused greater economic losses.With the development of Intelligent Transportation System,the technology of traffic condition judgement can identify road traffic conditions accurately and timely,providing decision basis for alleviating traffic pressure and improving traffic management.At present,the road network of large and medium-sized cities in China is very large.A medium-sized city can generate more than 10 millions pieces of passing vehicle data everyday.In such a large amount of traffic data environment,the traditional data mining algorithms are constrained by the performance of the single computer and cannot meet the actual needs in processing speed and other performance.The research on traffic state identification under the big data environment has become a new direction.After studying the classic algorithms in data mining technology,this paper first does data preprocessing operations on the measured traffic flow data,such as the deletion of redundant attributes,the elimination of abnormal data,and the normalization of data according to the traffic flow theory and the actual requirements of traffic conditions judgement.Under the normal urban order,the model between traffic flow data and road traffic status is a very complex dynamic problem.It has the characteristics of time-varying and nonlinearity.The strong self-learning ability of the neural network can make it fit the system to the greatest degree.This paper uses the single model approach to study the application of BP neural network,based on the actual traffic flow data which is collected by a city’s ITS.Under this research,the algorithm of traffic state identification based the multiple model theory is designed,which combines Fuzzy C-Means clustering analysis and BP neural network.The feasibility of this algorithm is verified by comparing the simulation results of single model and multiple model algorithms.For the environment of massive traffic data,this paper introduces the environment setup including Linux system,Hadoop platform,and Spark platform,and analyzes the algorithm implementation under the distributed platform.In face of big data environment,this paper uses the Hadoop and Spark platform with the powerful distributed computing capabilitie to implement Fuzzy C-Means algorithm and BP neural network algorithm.It explores the application of algorithms for massive traffic data,and analyzes the performance of the algorithm in different environments.The performance demonstrates the application advantages of big data platform in terms of operating speed.