Location Prediction Based On User Mobility Behavior In LBSN

With the rise of localization technology and the smart phones,Location Based Social Networks(LBSN)have gradually emerged,and a large number of users share their location information and life experiences through network platforms.As the number of users increases,these platforms have accumulated a large amount of data,and people get a lot of useful information by studying these data.Location prediction has also emerged.Predicting which users might visit in advance will not only bring convenience to users' lives,but also bring economic benefits to businesses.This thesis utilizes the data information generated by users in the location-based social network to study the location prediction of users in the social network.The research content are as follows:1.Based on the user's check-in location prediction at the new location,this thesis analyzes the distribution of the user's check-in space,time,and weather information,finds that the user at the new location has a similar movement pattern as the regular user,and proposes a location prediction model that integrates multiple factors.First,in view of the influence of spatial factors on the user's check-in behavior,a hierarchical KDE model is established based on the user's personalized behavior differences and the overall checkin behavior;then,the Markov model is used to model the inconsistency of the check-in time distribution;then,taking into account the constraints of weather factors on the user's mobile behavior,the "weather preference" feature is extracted for prediction.Finally,multiple results are linearly weighted to obtain the final prediction result.Experiments on real datasets prove that the model is superior to other baselines.2.This thesis also considers the impact of internal and external factors on the user's check-in,and predicts the location by studying the user's mobile pattern.First,the Apriori algorithm is used to mine the user's individual movement patterns to find out the internal factors that affect the user's check-in behavior;Then,we use DTW to calculate the similarity between the mobility patterns,and use cluster to obtain the overall mobile pattern of each group,which is the external factors that affect the check-in;Then,Markov model is trained based on mobile mode to predict the next location of users;Finally,considering the influence of external weather,we create weather collection to modify the results by calculating the weather similarity between the current location and other locations with Gaussian kernel function.Experiments on two cities of the LBSN dataset demonstrate the effectiveness of the model.