Research Of Personalized Event Recommendation In Event-Based Social Networks

Users are always unable to quickly and precisely find events with heterogeneous themes that best match their interests while they are searching for interesting events in event-based social websites equipped only with keyword matching and classified searching function.Event recommender system is targeted to solve this problem through mining users' event history and provide effective help for users,which is of great value in both theoretical research and actual practice.This paper proposes a personalized and contextual event recommender system that exploits overall user preference and context influences derived from users' historical events to produce recommendations in event-based social networks.To the best of our knowledge,we are leading the efforts in combining overall contextual evidence for the event recommendation problem.In summary,the contributions of this paper are as follows:1.We thoroughly analyze the scenario of users obtaining personalized event recommendations in event-based social service and the essential coldstart problem.2.We compute the semantic similarity between events through the training of the Latent Dirichlet Allocation model on events' contents,which speeds up the similarity computation,on the other hand,makes event recommendation more precisely.3.We compare and devise user preference matching measurement under multiple contextual factors from EBSNs including content preference,temporal impact,spatial constraints,cost consideration and social influence;This is one of the first attempts to leverage,collectively,the aforementioned contextual features for the event recommendation problem.4.We extract the aforementioned factors and evaluate their contribution to performance,and further combine them to overcome typical cold-start problem in EBSNs and make personalized event recommendation.5.We also design and implement this event recommender system in which server-side modules compute and predict user preference and process user requests.And we also apply the proposed approach to a real-world dataset called “Douban Events”,the experimental results demonstrate the effectiveness of our approach.This paper firstly introduced the related research background and reviewed literatures about event recommendation and their existing issues.After thoroughly analyzing the scenario of personalized event recommendation and the elements of an event,a personalized and contextual event recommender system that exploits overall user preference and context influences derived from users' historical events to produce recommendations is presented.At last,this paper described the design and implementation of this system,and the experimental results on a real dataset demonstrate the improvement of recommendation performance.