Event Recommender System Fused With Context And Attention Mechanism

In recent years,with the rapid development of online social platforms and the increase of the number of users,events and locations,it is difficult for users to quickly find their desired events or locations in a large amount of data.Therefore,the location and event recommendation of users in social platforms has been widely concerned and studied.Because event or location recommendation differs from traditional item or commodity recommendation in social platforms,precise location and event recommendation still face many challenges: in location-based social networks,user’s location changes over time,users’ long-term and short-term location preferences differ,and user’s preferences for both locations need to be considered in a comprehensive way.In event-based social networks,the number of users and events is much larger than the number of users participating in events(sparse data),resulting in poor performance of event recommendation algorithms.In view of the above two problems.Firstly,the user’s long-term and short-term location preferences are obtained from the context information such as location and time in user records using attention mechanism and cyclic neural network,respectively.The proposed algorithm is validated on the Foursquare dataset to effectively improve the recommended performance.Secondly,in event-based social networks,users’ preferences for events are obtained by using user-involved events and event context information,and the model is validated in the Meetup dataset to be superior to other algorithms.Finally,based on the event recommendation model proposed in this paper,a city event recommendation system is designed and implemented.The main work of this paper is as follows:(1)The authors presents an event location recommendation algorithm(LSTSRec)based on long-term,short-term,and time-space sequences.This algorithm makes full use of the location,coordinates and time information of the user’s check-in record.First,it captures the user’s short-term interest preferences from the user’s short-term check-in sequence using LSTM,and then captures the user’s long-term interest preferences from the long-term user interest sequence,distance difference sequence and time difference sequence using multihead attention mechanism.Finally,the algorithm is validated in the Foursquare dataset.(2)The authors proposes an intra-city event recommendation algorithm(ACARec)that integrates various context information and attention mechanism.The algorithm mainly integrates the context information of events such as event content,location,time,and group on the basis of user participation event data,and uses the attention mechanism to adjust the relationship between users and various context information.Compared with traditional algorithms,the proposed algorithm can more accurately recommend events to users.(3)According to the idea of intra-city event recommendation algorithm that integrates multiple contextual information and attention mechanism,a intra-city event recommendation system based on ACARec algorithm is designed and implemented by using the mainstream Web site development framework.According to the record of the user’s participation in the event,the system can realize personalized event recommendation for the user on the basis of comprehensively considering the content,group,time and location information of the event.