Semantic Analysis And Visualization On Big Data Of Urban Traffic Trajectories

In recent years,with the rise of concepts such as "digital city" and "wisdom city".GPS based sampling traffic trajectory data has been becoming increasingly popular as a research topic.Traffic trajectory data contains strong characteristics of space and time.Reasonable organization and display of such large data is quite difficult while doing research on traffic trajectory data.In addition,this paper developed distributed file system based on the framework of Hadoop,for the purpose of storing huge amounts of traffic trajectory data.And this paper also uses the Hive data warehouse to manage huge amounts of traffic trajectory data.According to the characteristics of the traffic trajectory data,in this paper,we adopt the concept of view and index in Hive to accelerate the searching speed and reduce the database retrieval cost.At the same time,we adopt the data partition based on the "day" and "hour" for the purpose of optimizing the storage structure of traffic trajectory data and reducing the times of full scan operations.In order to reduce the huge amounts of data's storage size,in this paper,we also attempt compression technology.In addition,this paper proposes a new way in research traffic trajectory data based on semantic analysis,which refers to the word vector model and programs using deep learning framework TensorFlow.In the visualization part,we calculate the traffic flow between city key grids and try to analyze the whole city traffic patterns.Then,we put forward some suggestions and opinions to improve the urban traffic.Finally,according to the research contents in this paper,we adopt an automated batch process system prototype based on Browser/Server structure,which includes functions of traffic trajectory data collection,pretreatment,trajectory data analysis and visualization display and so on.