Spatiotemporal Trajectory Clustering Of Moving Objects Considering Multidimensional Semantics

Spatiotemporal trajectory is a moving object's motion record in a certain time series.The spatiotemporal trajectory has rich temporal,spatial and semantic features.The spatiotemporal features and hidden semantic value of the spatiotemporal trajectory make the spatiotemporal track mining become a challenging frontier research branch in the field of data mining.Spatiotemporal trajectory clustering is one of the important research contents of spatiotemporal trajectory data mining.It aims to find the trajectory cluster with similar characteristics,which can be applied to trajectory prediction and anomaly detection.However,the data of spatiotemporal trajectory is large in quantity,and the data content and structure are complex.The essential characteristics of the spatiotemporal trajectory are hidden in the massive redundant information,and the processing procedure is complicated.In addition,it is difficult to analyze the temporal,spatial and semantic features of spatiotemporal trajectory data in a quantitative space.It is still a difficult problem to measure the semantic similarity of different trajectories.In view of the problem that it is difficult to extract and express semantic information hidden in spatiotemporal trajectory,this paper aims to consider the time,space and semantic characteristics of spatiotemporal trajectory data.Combining machine learning theory and technology,this paper studies the method of feature selection and the establishment of similarity measurement model of spatiotemporal trajectory.On this basis,the clustering of spatiotemporal trajectory based on semantics is realized,which solved the problem of tradictional geospatial clustering methods;By combining trajectory clustering with actual semantic environment,theoretical support for deep mining of highlevel semantics of trajectory data is provided.The main contents of this paper are as follows:(1)Spatiotemporal trajectory feature dimension reduction.The key information for specific fields and practical applications is usually hidden in high-dimensional and massive data samples.In order to quickly and effectively analyze the relationship between different spatiotemporal trajectories,it is often necessary to extract the main features from the original trajectories that can reflect their essential information.Trajectory data contains time,space and high-dimensional semantic features.Based on the highdimensional semantic features of spatiotemporal trajectory,combined with the practical significance of spatiotemporal trajectory research and the related theory of feature selection and feature extraction,this paper studies the feature dimension reduction method considering the semantics of spatiotemporal trajectory,and realizes the feature dimension reduction expression of high-dimensional and redundant spatiotemporal trajectory data.(2)Semantic similarity measurement of spatiotemporal trajectory.Aiming at the problem that the traditional trajectory similarity measurement only measures the spatialtemporal characteristics,a new method to measure the spatiaotemporal trajectory similarity is proposed,which considers the semantic features based on the feature dimensionality reduction of the spatiaotemporal trajectories.The multi-dimensional semantic representation is integrated into the trajectory similarity measurement to solve the problem that the traditional similarity measurement method can not measure the trajectory similarity at the semantic level.(3)Spatiotemporal trajectory semantic clustering.Aiming at the problem that traditional clustering methods mainly focus on the geographical spatial feature aspects and ignore the semantic features of spatiotemporal trajectory,a spatiotemporal trajectory semantic clustering method based on spectral clustering is proposed.This method integrates the semantic features of spatiotemporal trajectory,considers the similarity degree of spatiotemporal trajectory of semantic features,and combines the idea of spectral clustering with the actual semantic environment,to realize the spatiotemporal semantic information based clustering.Finally,the method proposed in this paper is verified and analyzed by experiments.In view of the similarity measurement of spatiotemporal trajectory,the comparison and analysis are carried out between different spatiotemporal trajectory features and existing similarity measurement methods,which verifies the rationality and effectiveness of the similarity measurement method that takes into account multidimensional semantics;In order to meet the application requirements of spatiotemporal trajectory clustering,the similarity results under different spatiotemporal trajectory features are selected to gather with traditional clustering methods respectively,and the validity and accuracy of the proposed method are verified after comparison and analyzation.