Research On Activity Arrangement Method Based On Directed Heterogeneous Graph In EBSN

Event-based Social Networks(EBSN)is a new type of complex heterogeneous social network.Users can publish or query events online and join offline.In recent years,with the rapid development of EBSN,how to arrange the optimal activities for users in EBSN according to their interests and preferences,that is,research on the method of event arrangement has become one of the hot issues.The research on the existing methods finds that the current activity recommendation algorithms in EBSN cannot be applied to activity arrangement because they ignore the conflicts between activities.Therefore,more and more researchers research the problem of activity arrangement in EBSN using data mining and other technologies.They analyze historical user behavior and predict user preferences to achieve activity arrangement in EBSN.Although the problem of too huge information has been solved to a certain degree.However,most algorithms ignore feedback from former users that can affect the quality of subsequent arrangements.Therefore,it has certain theoretical and application value to study the activity arrangement method with feedback in EBSN that can meet the needs of different users.This article studies the above issues in depth and divides the activity arrangement method in EBSN into two stages: offline preprocessing and online query.In the offline preprocessing stage,the activities,historical users and their relationships in EBSN are first abstracted into a directed heterogeneous graph.The nodes in the graph include active nodes and historical user nodes.In the order of occurrence,the edges of the historical user node and the active node indicate that the user has participated in the activity.The weight on this type of edge indicates the number of times the historical user has participated in the activity and the evaluation of it,which is expressed in the form of a dual.Secondly,in order to speed up subsequent queries and extract the attribute feature information of graph nodes and edges,a Directed Heterogeneous Graph Feature Index(DHGF index)is proposed.The index consists of node attribute feature indexes and directed edge attributesfeature index composition.In the online query stage,the query context is first converted into a query graph and the nodes and edges are filtered for the first time using the DHGF index.Secondly,a multi-attribute candidate set filtering strategy based on DHGF index is proposed,which utilizes the constraints of time,node ingress and egress,label type to further prune the query graph candidate set and avoid redundant calculation.For the new and historical users in EBSN,this article gives different query methods.Aiming at new users,an improved UCB activity arrangement algorithm with user feedback is introduced.Elastic net regression is introduced to calculate the user's interest value for activities based on multiple influencing factors and arrange activities with high interest values for it and receive whether users accept the activity feedback of.For historical users,this article proposes a method for arranging historical user events(HU_EA)by combining various factors.Use the meta-path method to calculate the top-k potential friends;use the number of activities and evaluations of historical users after preprocessing to achieve collaborative filtering;use the EN_UCB algorithm to calculate the user's interest value and integrate these three aspects to arrange the activity with the highest interest value for historical users and receive feedback on whether the user accepts the activity.A large number of experiments on real and simulated data sets verify that the method proposed in this paper which effectively implements online activity scheduling on EBSN.