Exploiting Multi-sources Information For Location Recommendation

The emergence of smart phones and the development of GPS technology have prompted the rapid development of location social networks.It can provide users with the ability to share locations,post comments and establish link with friends,which enriching people's lives.As the number of users increases,the amount of data increases exponentially.For example,Foursquare,as the most popular location social network,contains at least 50 million users and 10 million locations.Face to this massive amount of data,users cannot efficiently obtain the information they want.For this purpose and the personalized location recommendation system came into being.Personalized recommendation is the effective way to solve the problem of information overload.It serves as the core subsystems of application platforms such as social networking sites,e-commerce.and audio-visual sites that have achieved great success.The recommendation system analyzes the user's needs by modeling the historical behavior information of the user,and finally recommends the result to the target user through a recommendation strategy.In addition to being an important way for people to find interesting places,the location recommendation system can help merchants discover their potential users and achieve a win-win situation.It mines user preferences based on the user check-in data,comment content,geographical influence,etc.for recommending the locations to users,which can reduce their decision time.When exploiting the multi-source information to provide recommended services to users,it is also necessary to consider the diversity of user preferences,data sparseness,cold start and other issues.Based on the purpose of improving the performance of the recommendation system,this paper obtains user preferences based on the user's implicit feedback data,and integrates both geographical influence and social influence to accurately predict and recommend the locations to users.The research work of this paper is summarized as follows:1、We propose the new method called Base Item Attribute-Weighting Cosine Similarity(BIA-WCOS)that consider the influence of location attribute to social relationship to estimate user similarity.According to calculate the location's popularity and visits and fuses it into the cosine similarity as the influence weight,the proposed method makes the recommendation result more in line with real life and improves the accuracy of finding friends.The effectiveness of the algorithm is proved by experiments in real datasets Movielens.2、We propose a location recommendation framework that called Geographical-Base on the Location Attribute Social Relationship(G-BLAS)that combine multi-source data.BIA-WCOS is used to model the social relationship between users,which is considering the influence of location attribute to user similarity.Furthermore,fusing the user preference and location information to recommend the location to users.The effectiveness of the algorithm is proved by experiments in real datasets.