Spatial-Temporal Trajectory Data Minning

The development of mobile communication technology has led to a large number of spatio-temporal trajectory data generated by various industries.It has also spawned a large number of data mining studies on spatio-temporal trajectories.However,the research on trajectory mining is mostly concentrated in the fields of road traffic and so on.less.In order to extract the semantic information contained in the target motion trajectory in open space,this paper designs the trajectory vectorization method for the open space trajectory with less information,large data volume and high uncertainty.The coordinates are transformed into trajectory vectors containing semantic information to accomplish other trajectory mining tasks.In this paper,firstly,based on the high time complexity of density-based clustering algorithm,the grid-based improvement of DPC algorithm is proposed.The DPSCGNS algorithm is proposed,which can effectively extract clusters of arbitrary shapes and is robust to parameters.Time complexity is lower than other density-based algorithms.Secondly,the concept of trajectory action is proposed.The trajectory region serialization method based on hotspot region and the action serialization method based on trajectory action are designed.The annotation sequence that can be vectorized is obtained.Using a simplified neural network model,the Word2vec to embed discretized trajectory sequence to obtain a trajectory vector containing semantic information.Finally,using AIS ship dataset to design trajectory embedding clustering analysis experiment and trajectory target classification experiment,the feasibility of using trajectory vector for trajectory data mining task is verified,and the trajectory embedding method is proved to be effective.