Research On Personalized Spot Recommendation Algorithm Based On Location-Based Social Networks

With the development of the social networks and location-based services, location based social networks are increasingly popular. In this era of information explosion, personalized recommendation algorithms can effectively help social network users filter out information that they are not interested in. Using personalized recommendation algorithms also can explore users’ interests and greatly increase users’activeness in social networks. At present, recommendation algorithms have been widely applied on traditional social network systems. And recommendation systems are essential to location-based social networks too.Compared with traditional social network recommendation systems, location-based social recommendation systems can find user interests more effectively with the help of geographic location informations. At present, location-based social recommendation systems’ contents mainly include friend recommendation, activity recommendation and spot recommendation. The spot recommendation has become one of the most hot research areas. However, most spots in the location-based social networks are rarely checked in. What’s worse, some newly developed spots have none check in data. These make spot recommendation algorithms suffer from the sparsity of data and cold start problems.This dissertation studies the location-based recommendation systems and introduces the structure of location-based social networks, related recommendation algorithms as well as the common composition of recommendation systems. As there are three main factors that influence the personalized spot recommendation, this dissertation studies the three factors:user preferences, social influence factors and geographic location, and introduces three basic personalized recommendation algorithms:the user-based collaborative filtering algorithm, the collaborative filtering algorithm based on social influence and the naive bayesian classification. In addition, this dissertation deeply analyses the sparsity of data problem, cold start problem and check-in spots away from residence problem. And proposed an improved adaptive recommendation algorithm based on the analysis of the three problems. The improved algorithm combines user collaborative filter, social influence factor and naive bayesian classification, and it can adapt with users’current location for recommending the most suitable spot for the user.In this dissertation, we compared the improved algorithm with other recommendation algorithms through comparison experiments and verify the feasibility and effectiveness of the improved algorithm. The experimental results show that, the improved algorithm can effectively solve the problems in personalized spot recommendation and can recommendation better spots for users.