Statistical Modeling Of Multi-Source Traffic Data

The comprehensive utilization of multi-source heterogeneous data as an important research direction in the current big data analysis,has been widely concerned.This paper discusses two types of traffic data——text data traffic and traffic flow data,which has certain theoretical value and practical significance.This paper main work includes construction traffic text information extraction model,traffic flow time series modeling and visualization platform,specific work are as follows:Firstly,the basic theory of the conditional random field model is introduced.Based on the actual traffic text data,the Chinese word segmentation is completed,the traffic text annotation model and the traffic event element recognition model are constructed by using the conditional random field.The information extraction task is completed by using the model,The results show that the F values of the four elements of traffic text extraction location,vehicle,traffic condition and direction are 70.52%,87.50%,91.05% and 97.14% respectively,which are more than 70% and the effect is good.Secondly,this paper discusses the common time series and multivariate time series traffic modeling parameter estimation theory.On this basis,the introduction of interference analysis theory,the use of Guiyang new registered vehicle data model validation,the introduction of Guiyang limit line In the case of interference events,the model effect is improved remarkably,and the simulation of traffic flow data using multivariate time series analysis also achieves good results.Thirdly,using R software to achieve traffic data visualization.Through the traffic situation of the text classification,understands the occurrence of various types of traffic events.For more common concern traffic congestion,based on the use of map,you can clearly observe congestion is serious near Da Shi Zi of Guiyang central urban area.To sum up,this paper presents a comprehensive modeling scheme for the statistical modeling of the multi-source and heterogeneous data of the actual transportation department,and has obtained good effect and provided some help for the relevant departments.