Research On POI Recommendation Based On Collaborative Filtering Considering Temporal And Spatial Factors

With the development of the Internet and mobile terminal applications,nowadays more and more people would like to share different kinds of activities and information on mobile social networks.With the rapid development of positioning technology and location-based social networks,people would more like to share their location information in mobile applications.Therefore how to accurately recommend people to go the interesting places which they would like to go such as coffee shops,playgrounds?libraries(i.e.POI recommend)and so on is very important.Different from general item recommendation,time?space and social factor have an important influence on POI recommendation.If we do not make full use of these factors,the accuracy will be very low.To solve the problem,the paper first introduces the collaborative filtering technology.Collaborative filtering technology on POI recommendation is to computer the similarity between users or POIs,and then recommend POIs that people had not visited before to people based on the similarity.At the same time,the influence of geographical position factor is integrated.The paper fit the POI probability-distance in power law distribution model,and then computer the coefficients of the model with the least square method.We propose a fusion algorithm which considers the geographical factors based on user-based collaborative filtering algorithm.Then we consider the influence of time factor.We take into account of similar users' time and user's historical time influence.In the end,we consider the user preference factor the geographical factor and time factor to build a recommendation unified model to calculate the recommendation score to recommend POIs to users more precisely.We examine our algorithm on Foursquare and Gowalla data sets to compare with existing excellent recommendation algorithms on POI recommendation.The results show that our algorithm has a certain improvement in accuracy and recall rate compared to present algorithms on POI recommendation.