| In recent years,location-based social networks(LBSNs)that connect users’online and offline social interactions,such as Foursquare,Gowalla,WeChat and MicroBlog,have become increasingly popular.They provide a convenient platform for users to share their activities with families,friends,and classmates,and each activity in LBSNs is associated with some attributes,including the user participating in the activity,the timestamp of the user check-in,the location of the activity,etc.With a large amount of information in LBSNs constantly being checked in,it is difficult for users to find venues of interest from the vast data.To address this issue,venue recommender systems have emerged.Previous studies mainly focused on personalized venue recommendation,while there were relatively few studies on group venue recommendation.At present,the accuracy of group venue recommendation is low,and there are issues such as data sparsity and cold start.Therefore,in response to these issues,this thesis conducts in-depth research on group venue recommendation approaches in location-based social networks,aiming to provide more accurate and efficient venue recommendation services for a group of users in LBSNs.The main research content and innovation of this thesis are summarized as follows:(1)Multi-pattern group venue recommendation based on time awareness.In response to the issue that existing research has not yet considered recommending venues to a group of users based on different temporal patterns,by analyzing the temporal pattern of group visit venues,it is found that group activity venues have a time effect,that is,the venues visited by groups are different under different temporal patterns.Based on this discovery,a multi-pattern group venue recommendation approach based on time awareness and an improved grouping algorithm are proposed.The experimental results on two LBSNs datasets show that using this approach can accurately capture the venue preferences of groups in different temporal patterns,and can more accurately predict which venue is more likely to be visited by the group when in different temporal patterns,thereby improving the accuracy of group venue recommendation.(2)Group venue recommendation based on time and content dual topics.Regarding the existing group venue recommendation approaches,which mainly rely on single topic modeling and have serious data sparsity problems,it is found through analysis that there are latent relationships among the time,content,venue,and participants of group activities,especially the correlation between the activity time and the activity content,that is,the activity content that the group participates in at a certain time is often similar.Based on this discovery,a group venue recommendation approach based on time and content dual topics is proposed.The experimental results on two LBSNs datasets present that this approach extracts the time and content topics of activities from the perspective of joint modeling of time and content dual topics.It can more accurately obtain the group’s preference for activity venues,alleviate the problem of data sparsity,and improve the quality of group venue recommendation.(3)Group venue recommendation based on geographical location and activity content awareness.In view of the existing group venue recommendation,which mainly focuses on spatio-temporal awareness and has the problem of venue cold start,it is found through analysis that geographical location plays an important role in location-based social networks,and there is a latent relationship among the geographical location,activity venue,and activity content of group activities,especially the correlation between the activity venue and the activity content,that is,the activity venues selected by groups for the same activity content often have similarities.Based on this discovery,a group venue recommendation approach based on geographical location and activity content awareness is proposed.The experimental results on two LBSNs datasets manifest that this approach not only alleviates data sparsity and venue cold start problems,but also outperforms the current mainstream group venue recommendation approaches in terms of recommendation performance.(4)Multi-awareness and multi-topic group venue recommendation based on multi-source information fusion.Aiming at the diverse needs of group members for decision-making activity venues,it is found through analysis that locationbased social networks contain abundant available spatio-temporal semantic information,and group decision-making venues involve multiple domain topics.Based on this discovery,a multi-awareness and multi-topic group venue recommendation approach based on multi-source information fusion is proposed.The experimental results on two LBSNs datasets indicate that this approach not only describes the group activity behavior more comprehensively,but also meets the diverse needs of group members and improves the accuracy of group venue recommendation by fusing multi-source spatio-temporal semantic information of members,content,time,venue and location in group activities,and using the topic probability model to automatically extract the content topic,time topic and regional topic of activities. |
