Research On Location Prediction Over Check-ins

Location-based social networks(LBSNs)is the integration of positioning technology with traditional social networks.LBSNs allow users to publish text,pictures,and video.At the same time,you can also share the current location,which together constitute the Check-in.LBSNs' s more rich social experience to attract more and more users,a large number of users in the use of the process to generate a massive sign data.Due to the location information,the check-ins can reflect the user's mobile behavior.Position prediction is an important research direction in the field of human mobile behavior research.Because of the easy access and the large scale of the data,many related studies have been used as data sources in recent years.The main work of this paper is as follows:(1)A comprehensive analysis of the existing research.Analyzes three kinds of position prediction models,temporal-based,spatio-based and temporal-spatio-based.Then studies the position prediction model of social relations.(2)Analyzed the characteristics of Check-ins from temporal,spatio and social.Analyzes the regularity of the check-in time and the regularity of the check-in point from the perspective of the whole and the individual user respectively,and examines the influence of the social relationship on the user 's check-in behavior.(3)The temporal-based prediction model has been improved.When the relevant probability is calculated by the Gaussian mixture model,a specific method to determine the number of samples is the number of Gaussian components,and the calculation model suitable for the user can be established according to the checkout feature.(4)The spatio-based prediction model has been improved.In view of the shortcomings of the traditional Markov model,this paper proposes a spatial prediction model with sequence similarity instead of the transition probability,which takes full account of the influence of historical information on the user's next position.(5)Proposes a prediction model based on temporal and spatio for the introduction of social relations,and the improved temporal-based and spatio-based model is combined with the influence factor to improve the accuracy of prediction.It also effectively alleviates the data sparseness problem.Experiments on the Gowalla check-in dataset show that the predictive accuracy of the proposed model is improved compared with the traditional and state-of-the-art approaches,and good results are obtained.