The Research Of Personalized Recommendation Technology Based On Tag And LBS

With the rapid development of mobile Internet and mobile communication network and mobile intelligent terminal equipment increasingly popular, the original PC service and platform also gradually shift to mobile terminal. In the mobile network environment, the user’s mobile is recommended demand greatly influenced by the mobile situation, accompanied by a mobile terminal screen, poor capacity of small, input speed is limited, and many other factors, the user have higher request for recommended recommend real-time and accuracy. Dimensional in mobile environment, traditional recommendation methods can not meet user requirements.For recommended requirements in mobile environment, this paper proposes a combination of LBS and social networks tag recommendation methods. With the help of the social tagging behavior of users in a mobile environment and the trust of the existing social network relation, from the users of high correlation with the target users looking for information on reflect users’ interests; At the same time considering the user’s interest in mobile environment change with time and space, the introduction of user’s location and mark time, get more in line with the user interests of personalized information, improve the recommendation of the real time information, recommended to reduce the blindness of, thus improving the accuracy of recommendation.Recommended method principle is:first, using the existing positioning technology for user’s location information, and geographical location according to the pyramid model hierarchy and area; Then, the application of social network analysis method to analyze user-resources-label, for users with resources and social relations of double layer network model; Comprehensive hobbies and space position again, to get double model integration; Fusion are calculated separately, and the double layer model of user similarity relation similarity and user space position; Finally, the two weighted similarity, get the user’s comprehensive similarity, high project integrated similarity to recommend item set, returned to the target user program is recommended.In this paper, comparative experiment on Foursquare data set. The experimental data including personal information, label information, social relations and the placement sequence; Contrast algorithm including the recommendation algorithm proposed in this paper, the traditional collaborative filtering algorithm and recommendation algorithm based on LBS. The recommendation of the experimental results show that the proposed algorithm is effective and feasible, in terms of precision and recall rate than other two algorithms are improved significantly; And, when increasing amounts of data, this paper puts forward the recommendation algorithm has good scalability, at the same time can effectively avoid the cold start problem.