Clustering Analysis Of Taxi Spatio-temporal Data Oriented To Region And Track

With the development and progress of science and technology,people’s living standards have improved significantly,and the social problems are also very sharp.Followed by the the large increase of traffic vehicles,the traffic congestion is very serious.We must try our best to find method,which can solve this problem.At the same time,because of the highly developed technology,a variety of sensor equipment and satellite positioning equipment make position and navigation for people.They also recorded a large number of vehicle mobile data.These massive traffic data contains a lot of information.Through these data,we can find the law of urban traffic congestion and the law between urban hot spots.However,all of this information needs us to discover.At present,most of these traffic data are positioning information provided by GPS system.In many cases,these information is not only the longitude and latitude coordinates.It also contains many other information,such as time information and the speed of car.The taxi data set is the most easily accessible and representative data in different kinds of data sets.His movement in the city has a great randomness compared with the bus,and has a long continuity in time compared with the private car.This is very worthy of analysis.At the same time,face the problem of urban congestion,congestion in a region is not only a problem at a certain time,but also a general trend in a period of time.As a result,many clustering algorithms on the plane are not suitable for the current situation.We need to extend the applicability of clustering algorithm to achieve spatial and temporal extension.The spatio-temporal trajectory with time information is used to complete the congestion judgment on spatio-temporal.In order to deal with this problem,various data mining algorithms emerge as the times require.This thesis will use a series of density-based clustering algorithms to analyze the spatial and temporal characteristics of traffic congestion in the region and in the trajectory.Firstly,ST-DBSCAN clustering algorithm will be improved.This algorithm introduces the time dimension,expands the traditional clustering algorithm from plane to a cylindrical space-time volume,and deepens the spatial-temporal relationship between clustering objects.At the same time,a series of non-spatio-temporal constraints are introduced to optimize ST-DBSCAN’s decision of congestion area.The main constraints introduced are speed constraints,time constraints,frequency constraints of vehicles in space-time,and direction constraints of vehicles.Then,the spatial and temporal characteristics of traffic congestion in the region are analyzed by clustering results.Secondly,the TRACLUS algorithm is improved.Based on the traditional TRACLUS algorithm,the time dimension is extended,and a spatio-temporal trajectory clustering algorithm ST-TRACLUS is obtained.In clustering,a sub-trajectory segment division algorithm with minimum description length is introduced,and the original trajectory points are divided into a representative sub-trajectory segment.The object of clustering algorithm is transformed from trajectory point to sub-trajectory segment.At the same time,a series of non-spatio-temporal constraints are added to the clustering,and the optimal algorithm is used to determine the congestion trajectory segment.Finally,the spatio-temporal characteristics of traffic congestion on the trajectory are analyzed by using the results of spatio-temporal trajectory clustering.