Research On Social Relations Inference Model Based On Spatio-temporal Data

With the development of positioning technology and communication technology,the tracking accuracy and accuracy of moving objects are improved significantly.With the development of mobile intelligent devices and computer storage technology,it is more convenient to acquire location data of mobile objects,and the data can be stored permanently.A large number of historical location data provide abundant data basis for the behavior analysis of mobile objects.As an important part of daily life,social relations are widely used in sociology,public safety,pervasive computing and product marketing.In this thesis,people are regarded as mobile objects,and the trajectory data generated by people's daily activities are studied.From the new perspective of space-time trajectory data,a social relationship inference system is designed and implemented.The main work of this thesis is as follows:(1)This thesis summarizes the existing algorithms of residence point extraction in spatio-temporal data mining,and studies the problems of "pseudo residence point" and "false segmentation of residence point" in existing algorithms.Based on clustering analysis,the idea of DBSCAN algorithm is selected as the algorithm basis for residence point extraction.The input parameters of the algorithm are improved,and the judgment of candidate residence area is added to the algorithm,which improves the adaptability of the algorithm to complex scenes and lays a solid foundation for subsequent research.(2)A method of transforming spatio-temporal data into semantic trajectories is proposed.By introducing POI database to add rich spatial scene information to trajectory data,the implicit information in trajectory residence points can be extracted better.From the perspective of semantic trajectory,the multi-level spatial scale semantic trajectory is established according to different spatial granularity to achieve efficient acquisition of co-occurrence data among users.(3)Based on trajectory data,a social relationship inference algorithm supporting probabilistic output is proposed.Considering the complexity of social relations in real life,we can get the inference results of social relations and give the probability of belonging to the category of social relations.Compared with other inference methods of social relations,we can present more decision-making information.(4)A visual inference system of social relations is implemented.Combining with the spatial scene information,the user's trajectory data is converted from abstract longitude and latitude data to visual or semantic information,and the visual query of trajectory data is realized.Based on the trajectory information,the inference function of social relations is completed,and the topological graph of multi-person social relations is displayed.The test of the system shows that the system meets the expected requirements in function realization,and has friendly interface and correct function.