Comment Analysis Of POIs Prediction Method Based On LBSN

In recent years,with the continuous development of social networks,more and more users use various social networks to share their preferences for something or some pointsof-interest(POIs).A large number of users have a check-in,comment or share their POI via WeChat,Foursquare,sina weibo,dianping and other social app.At the same time,the popularity of smart devices will enable more users to check-in through social apps and people can view,thumb up,and comment other interested content posted by other social users.Social platform with the increase of users,the users sign in data storage is growing exponentially.Through these data to predict the user's next POI is becoming more and more important,with the forecast of user's POI,users can enjoy more and more personalized services,and business can target customer service accurately and achieve good profits.In this thesis,we study the POI prediction in social networks by using the comment information of the user's check-in location.The main results are as follows:1.Based on the research on the user check-in data,this thesis proposes a POI prediction model based on the comment information in the user check-in data.First,according to the user's check-in data,we extract users' intention of POI with the use of tensor decomposition,and then classify each user's sign in position with the use of time and POI features to classify each user's check-in in position,and each user's check-in can be expressed as an intention.The hidden markov model is used to predict the user's next check-in intention.At the same time,with the considering of the spatial distance of users' continuous check-in positions,we narrow the prediction range.Furthermore,we obtain user's comment through user history check-in and the POI's comment can be accessed under consideration of this intention.We use the topic model to extract the topic preference to obtain the distribution of the words,and we use JS distance to obtain the topic similarity between the two documents,and get the POI prediction result of top-k.The experimental results show that the strategy proposed in this paper is effective and improves the prediction accuracy and data scalability.2.We are trying to solve the problem of slow computation due to the large amount of data in the POI prediction of the topic model,this thesis proposes the method of using spark for distributed computing.Compared with single machine operation,the parallel operation can improve the operation efficiency.