Design And Implementation Of Point Of Interest Recommender System Based On Social Network And Check-in Data

With the development of the Internet,location-based social network(LBSN)services have been widely used,and how to help users choose useful information from the exponentially explosive growth of information has become an important research.Compared with the traditional recommender application scenarios,there are more complex contextual information in LBSN.These information are closely related to the user's travel check-in behavior.Making full use of these contextual information for modeling to better simulate users' travel decision-making behaviors to improve the quality of recommendation results of points of interest is a problem that requires in-depth research.Based on this,the main research content of this article has the following parts:(1)This paper proposes a new STSTM(Sequential Temporal-Spatio Topic Model).STSTM can explore potential themes and potential regions that change over time.That is,the time factor is the upstream factor that affects the subject and the region.STSTM uses continuous time instead of discretized time as a model.In order to reflect the influence of previous check-in selections on the next check-in selection,STSTM integrates the timing impact of each check-in into the model.By using the Latent Dirichlet Allocation(LDA)model to mine the probability of potential topics,time intervals,and potential regions,combined with predicting the timing impact of the check-in location,a prediction list for the next check-in is generated.This paper conducts experiments on a real LSBN sign-in data set,and the results show that this algorithm has a better recommendation effect than other advanced interest point recommendation algorithms.(2)This paper proposes a Temporal Joint Model(TJM)based on time mode to simulate the check-in behavior of users in the decision-making process.This model strategically integrates the above factors and makes full use of them.In order to prove the applicability and flexibility of this model,we studied how it uniformly supports two recommendation modes,namely weekday recommendation and holiday recommendation.We conducted a large number of experiments on two real large-scale data sets,and evaluated the performance of the model in terms of recommendation effectiveness and recommendation efficiency;(3)In order to take advantage of the feature of check-in comments,this paper proposes a topic model(SAETM)based on user social networks and sentiment analysis of comments.SAETM can capture user interests in check-in behavior from two perspectives:user social networking and comment analysis.First,use the latent Dirichlet distribution model to mine text information related to points of interest;then build social relevance through user social relationships to generate social relevance scores.Finally,the probability model is used to effectively integrate the relevant scores of geography,time,society,and interest,so as to generate a recommendation list recommended to users.This article is based on a real LBSN sign-in data set for experiments,and the performance of the model is evaluated from the effectiveness of the recommendation.The experimental results show that it has a better recommendation effect than other points of interest recommendation algorithms;(4)Based on the proposed three points of interest recommendation models,this paper designs and implements a point of interest recommender system.The system integrates a variety of personalized recommender algorithms,which can combine the user's check-in data to provide users with recommendations of points of interest in different check-in scenarios to meet user needs and improve user experience.