Research And Application Of Point Of Interest Recommendation Based On Neural Network

With the rapid development of mobile social networks,smart phones have become the main mobile devices for people to obtain information and network communication.Users can use phone to share their location with their friends through check-in.Mining those check-ins could provide users with more excellent services such as point-of-interests(POI)recommendation that is an importance part of mobile recommendation system.POI recommendation aims to help users find interesting locations they have never been and it will bring potential advertising revenue.Hence,POI recommendation has attracted more attention from academia and industry.POI recommendation provides personal services for users.At present,there are plenty of researches about that but most of them are based on traditional algorithms which consider of much additional information of mobile environment.Therefore,they cause sparsity problem of check-ins,which make similarity computing between users more difficult.To solve the problems mentioned above,this paper proposes a deep learning model based on user check-ins data.The main work is as follows.(1)Introduce the background,research significance,status of POI recommendation and analyze shortcomings of traditional algorithms.(2)To compute similarity between users by check-ins data,a deep learning model(POPDNN)that takes location popularity into account is proposed.Experiments on Gowalla dataset prove POPDNN perform better than traditional algorithms.(3)Propose a POI recommendation model(RBMNMF)that takes two-way analysis between users and POI.Use neural network to mine potential user-POI relationship which incorporate nonnegative matrix factor(NMF)for improving the smoothness of model.Experiments on Foursquare dataset prove RBMNMF has an excellent accuracy when recommending.(4)Design a prototype of scenic spot recommendation system called Travel Go that consists of many functions such as check-in activity,photo sharing,POI recommendation and so on.The system aims to recommend locations for users by mining their check-ins data.