Analysis Of Spatio-temporal Integration Features And Prediction Of User Trajectory Data

With the widespread use of mobile devices and the rapid development of positioning technologies such as 5G,location-based services have gradually become an indispensable part of human life styles.More and more user trajectory data are passed through various movements.The equipment is collected,and the spatio-temporal context trajectory data contains a large amount of potential user behavior information,which is essential for user behavior pattern mining and next footprint prediction.It is one of the research hotspots of artificial intelligence in recent years.This dissertation proposes a trajectory prediction model based on multi-scale feature fusion and adaptive clustering,using this model to mine and analyze user behavior patterns,so as to realize the user’s next footprint prediction.The model incorporates the three algorithms proposed in this dissertation for trajectory cutting,spatio-temporal feature extraction and trajectory prediction.Algorithm 1 is an adaptive timestamp definition algorithm based on time statistics.This algorithm combines the idea of statistical analysis to determine personalized adaptive trajectory cutting timestamp for each user,so as to solve the problem that traditional methods introduce subjective factors and ignore the differences among users;Algorithm 2 is a multi-method integrated feature extraction and normalized cause-and-effect embedding algorithm.This algorithm introduces the time series feature extraction method and focuses on the potential feature information of trajectory data as a whole,which solves the problem that traditional methods only focus on a single feature space and ignore the dimension of feature comparison;Algorithm 3 is a multi-scale feature fusion and adaptive clustering algorithm,which solves the problem of information sparseness caused by personalized user check-in behavior through common pattern mining.In this dissertation,the proposed hybrid model is validated in a real data set,and the experimental results show that the hybrid model is superior to the existing comparison methods.At the same time,the proposed algorithm and hybrid model are applied to the trajectory data management and trajectory prediction in the security field,which lays a foundation for the realization of security decision.