Research On Personalized Rcommendation Algorithm Based On Clustering In Mobile Environment

With the micromation of mobile devices and the development of wireless communication technology, GPS positioning technology and the electronic sensors, in mobile environment, location-based personalized recommendation algorithm has aroused extensive attention. Recently,as the rapid increase of mobile users and the explosive growth of information in network, which gives a great challenge on the scalability and the real-time of the current recommender system. At the same time, since the variability of the location and environment of mobile users, which requires the recommender system to respond quickly on the changes of users'interest and provide satisfactory recommendation to users. This paper mainly researched on location-based personalized recommendation algorithm in mobile environment.Firstly, aiming at the problem of the scalability and the real-time performance of recommender system in mobile environment, we proposed an user clustering-based collaborative filtering algorithm (UC-BCF). Which separates the procedure of recommendation into offline and online phases. In the offline phase, the data of mobile users are preprocessed, and the mobile users are clustered based on the distance between users. While in the online phase, the algorithm only searches the most similar neighbors from the clusters whose cluster centers are nearer to the location of the mobile user, and produces recommendation.Secondly, aiming at the problem that the system need to changed quickly according to the varied of user'interest, we proposed an item clustering-based personalized recommendation algorithm. The algorithm uses K-means clustering technology, case-based user model and multi-attribute decision making method. During the process of recommendation, the algorithm determines the weight value of each attributes according to the user's attention on each attribute, improves the attributes standardization formula and the weight formula. Which makes the recommendation algorithm can deal with user's short time interest.Finally, experimental schemes for user clustering-based collaborative filtering recommendation algorithm and item clustering-based personalized recommendation algorithm are given, the response time, recommendation accuracy and the interest sensitivity of the two algorithm is verified respectively.