Research On Collaborative Filtering Recommendation Algorithm For Mobile Users

As the development of Mobile Internet, the problem of information overload has been accelerated. People can not accurately and timely access to their information needs. In the face of information overload, the recommendation algorithm is the best solution to solve the information overload. Recommendation algorithm after the long-term development, based on the traditional PC recommendation system has been mature, and mobile recommendation research is currently in the initial stage. For the traditional PC environment, mobile environment with richer information characteristics, such as the context information, has location information and time information. Compared in traditional recommendation, score information, the position information and the time information of the mobile user is difficult to establish data model. The research about how to make full use of the characteristics of these information to bring more accurate and timely recommendations for mobile users is urgent for mobile user recommendation algorithm.At present, the research of mobile recommendation algorithm is being widespread attention at domestic and abroad. However, most of the literature is based on the single-dimensional or static-based mobile recommendation algorithm, The information in mobile environment is multi-dimensional and dynamic. It’s the comprehensive effect of location, time, social relations and individual users preference. For multi-dimensional and dynamic analysis, In this paper, the location information of mobile users and time effect information has been analysis. Proposed a collaborative filtering recommendation algorithm based on multidimensional dynamic information. Combine location, time, and user’s item score for 3D information with Collaborative filtering recommendation,to achieve a mobile recommend with more suitable for the environment of Mobile Internet. Specific work is as follows:1) Proposed a description of location information of mobile users and according to the location information of the users, symbolic modeling is carried out. According to the characteristics of Mobile Internet connection and the position information service, formulated the scoring mechanism of mobile users, achieve the location of the mobile user rating assumptions, which has been the location of the user ratings matrix. Then introduce the Pearson correlation coefficient analysis of the similarity of user location. The algorithm idea is provided for collaborative filtering of location information. Combined with the traditional collaborative filtering algorithm, achieve the accuracy of mobile users recommended.2) According to the time characteristics of traditional Internet users’ browsing time and the time effect of mobile users is different, The time effect of mobile Internet is summarized. Mainly reflected in four aspects, respectively is: persistence of interest, attenuation and the epidemic, seasonal. And then the four time effects are modeled and analyzed. And finally combined with the traditional KNN recommendation algorithm, a collaborative filtering recommendation model based on time effect is constructed. The proposed algorithm solves the timeliness and dynamic of mobile recommendation.3) The position information collaborative filtering algorithm and the time effect collaborative filtering algorithm are fused in this paper, is proposed for mobile user collaborative filtering algorithm. The algorithm solves the multidimensional information analysis and dynamic recommended. The recommendation for mobile users is more accurate, regional, timely.4) Through experiments and the comparison of several typical recommendation algorithms show that the feasibility of multidimensional dynamic collaborative filtering recommendation algorithm for mobile users is proposed in this paper. The results show that mobile recommendation algorithm integrating location and time is more suitable for the recommendation of Mobile Internet environment.