The Research On The Mobile User Behavior

Recently,With the rapid update of Mobile terminals, the efficiency of mobile users labeled and uploaded their positions through cell phones and other mobile terminals gets rapid ascension,we can easily get our location data at any time with high accuracy. At the same time,all kinds of terminal equipments have accumulated vast mass amounts of user data,which contains mobile users’ behavior patterns, interest preference information and other valuable information. This issue we study on the point of interest annotation data of mobile users, focus on the point of interest information in the trajectory, to analyze point of interest semantic and behavior, the results of the analysis applied to find user behavior similarity and mobile user location prediction.First, we find users’ behavior similarity from the labeling point of interest data and processing the information to semantic normalization.Research to do the following works:(1)find users’ point of interest. According to the characteristics of users’ mark points of interest,establishes users’ interest semantic ontology tree to collect semantic POI.(2) semantic ontology mapping. We calculate similarity between the users’ POI information and ontology tree node by semantic similarity calculation method.An experiment is performed to evaluate the performance of the new similarity measure by using the trajectories of twenty users in a period of five months,the results show that the method has high accuracy.Second, combine with the ontology semantic tree using Markov model to analyse the users’ location. Because of the complexity of the mobile users’ behavior,only consider the label number of the POI and the label time is not enough.Therefore,this paper proposes a new prediction method based on markov model to caculate jumping probability of mobile users’ semantic POI.Research to do the following works:(1)establishes semantic POI markov model.(2) proposes a new calculation method to improve the jump probability calculation,and then compared with standard markov model and second-order markov model calculation to verify the accuracy of the improved method.