Point Of Interest And Route Recommendation System For LBSN

With the rapid development of web 2.0,mobile positioning technology and the popularity of smart phones,a large number of location-based social networks have emerged in recent years,like Foursquare,Facebook Places,and Gowalla.Unlike traditional social networks,LBSN has introduced location tags that allow users to leave their footprints in social networks anytime,anywhere,and share their loved points of interest and travel experience with their friends.And with the rapid development of online shopping and social networks,the sparseness of data has intensified,and the quality of recommendation of traditional collaborative filtering methods have reduced a lot,sometimes even failed to get recommendation results.Dishonest users can forge a large amount of user rating information to change the recommendation result,lead to the collaborative filtering method is more vulnerable to malicious attacks.Moreover,the time cost of CF-based method is associated with the number of users and items.As the number increases,the performance of the methods decrease a lot and the scalability is poor.Therefore,traditional collaborative filtering methods urgently to introduce new data sources to improve algorithm performance and recommendation quality.On the other hand,when a user visits a new city,although there are some travel guide websites that can provide a lot of content,such as attractions photos,reviews and detailed travel itineraries.However,it could be a time-consuming work to obtain useful information from raw materials that have not been processed.Therefore,automated and personalized recommendations are highly anticipated and loved by users.In particular,personalized recommendations have attracted a lot of attention as they can effectively integrate the user's personal preferences(such as culture,personalized preferences and habits.),and give users high-satisfaction travel experience.In response to the challenges and problems in POI recommendation and trip recommendation research,this article studies from the following two aspects:(1)This paper integrates the trust relationship into the POI recommendation system.On one hand,the trust relationship in social network can reflect the similarity and influence between users;on the other hand,adding the trust relationship can effectively solve the cold start and malicious recommendations problems in traditional methods.This paper analyzes the propagation characteristics of trust and distrust relationship,and gives representation and compute method of trust relationship.And proposes a hybrid recommendation system that combines user similarity,geographic location and trust relationship.(2)In this paper,we use user's historical travel records to mine their personal preferences,and looking for high-satisfaction travel routes which satisfying the user's time and cost constraints.We proposed the TripPlanner recommendation system in this paper,first we construct a POI scoring model which based on user and time characteristics.Then,we present a hybrid trip mining algorithm which based on state expansion,and it guarantees to find an optimal and personalized route.In addition,we also proposed two pruning strategies to improve the efficiency of trip mining algorithm.