Efficient Retrieval Of The Top-k Most Relevant Event-Partner Pairs

The flourish development of event-based social network(EBSN)attracts a lot of attention from the research community.EBSN uses events to combine online and offline network.It includes online publishing events,appointment registration,and offline participation.Users can not only make friends online,but also participate in the event and make friends offline,which is not available in traditional social network.Due to the complexity of online and offline network,the types of information are very rich.Many research efforts have been spent on event recommendation,friend recommendation,location recommendation,and the recently emerged event-partner recommendation.However,these recommendation algorithms have high computational complexity and long-delayed.Moreover,it cannot meet the users'requirement to pull the information.Although the information retrieval algorithm allows users to take the initiative,it only returns results related to the query keyword.The type of result is single and cannot meet the user's need in practical application.In addition,users are likely to give up participating in the event if there is no partner.In order to make up for these shortcomings and realize the importance of a partner for a user to attend the event,we combines the advantages of information retrieval algorithm and recommendation algorithm.This paper propose a new information retrieval model:the top-k event-partner(kEP)pair information retrieval query model.This paper includes the following aspects:(1)Propose a new kEP query model.The query user inputs the keyword to search event.Then the algorithm returns the related events and pairs each event with a best partner.In other words,it returns the top-k event-partner pairs to the query user.(2)Based on kEP query model,this paper propose an efficient computing framework.(1)Design the event-user bipartite graph and inverted index?_d to quickly retrieve the data.(2)Set a threshold to terminate the algorithm early.(3)Borrow the idea of rank join,which sort when calculating.Design and compare the two join strategies:nested loop join and ripple join.(3)In order to efficiently computing the results of kEP queries,this paper proposes three pruning optimizations:(1)the unpromising event pruning technique can eliminate the events that cannot become the results.(2)the key partner technique can quickly find event-partner pairs.(3)the technique of efficient partner computation can quickly find a best partner for an event.(4)Crawling a large amount of real data in EBSN.The experimental results in the real data set shows that the proposed optimizations improve the performance of the framework significantly.In most cases,the ripple join is better than the nested loop join.