Design And Implementation Of Group Events Recommendation System Based On Context-aware And Multiple Features

With the rapid development and diversification of Internet products and services,event-based social networks(Event Bsaed Social Networks,EBSNs)have also developed rapidly.A large number of various events are frequently generated on the EBSNs social platform.It is often difficult for users to choose suitable eventsfrom these events.Therefore,the event recommendation problem of event-based social networks has been widely concerned and studied.However,accurate event recommendation for users still faces many challenges:Most of the eventsin EBSNs are jointly participated by some users in group events,so groups of users with similar interests will be generated.When recommending eventsfor group users,you need to weigh the group Different preference characteristics of each member;the content of eventsand rich attribute characteristics are important factors affecting user decision-making.However,traditional group recommendation rarely considers multiple characteristic factors comprehensively to mine the comprehensive influence of multiple scenarios on group preference from the perspective of the group.In order to solve the above problems,this paper proposes a new group event recommendation model based on the context information in EBSNs and the group's behavior preferences for events.It is verified on the Douban dataset that the model effectively improves the recommendation Performance.Finally,based on the model proposed in this paper,an intra-city event recommendation system is designed and implemented.The content of this article is as follows:(1)A group event recommendation model(TGRec-kernel)is proposed which combines label features and kernel functions.This model mainly uses kernel functions to perform high-dimensional semantic expansion of group and event label features,improves the similarity calculation method of group label preferences and event label features,and can more accurately model group-to-group eventscompared to traditional models Interest preferences.(2)Propose a group event recommendation model(FN-PMF)for mining interested friends based on target social relationships and geographical factors.This model makes full use of the social relationships of group members and the geographical location of participating events to mine interested friends,uses the interested friends to generate candidate events to model local preferences of groups and events,and uses a matrix decomposition model to globally build group preferences.The model is verified on the Douban data set,and the model can get better recommendation performance.(3)A group recommendation model(DP_GPPMF)based on dynamic behavior preference in the spatiotemporal context is proposed.The model is based on the TGRec-kernel model,which takes into account changes in the group's preferences for event and tag characteristics over time,event time,and geographical location characteristics,and other group dynamic behavior preferences.The model is verified in the Douban dataset for accuracy and normalization Cumulative damage gain rate has been significantly improved on the two evaluation indicators,and it can alleviate the cold start problem of the event to a certain extent.(4)Design and implement an intra-city event recommendation system based on the two models of FN-PMF and DP GPPMF.The system includes event information and user function modules.It recommends events by extracting feature data such as tags,time,geographic location,and social relationships,and uses interest tags to alleviate cold start problems for users.The system uses the mainstream web site development framework and Restful design style for design and implementation,and tests the function and performance of the system.