Research And Implementation Of Social Recommendation System Based On Diversity

With the rapid developing of mobile phone service, Telecommunication Carriershave collected a lot of user purchase information. Mean while, they have more and moreinformation content needed to be pushed to users. In this process, user preference inpersonality plays an important role, while the traditional marketing method is easy tocause the users feel uncomfortable or even disgusting. In this case, for each userprovided a personalized marketing strategy is very effective. In this thesis, we designedand implemented an information recommender system. The basic principle is accordingto the user’s purchase records and social behavior information, by using the data miningand social network analysis method to generate personalized marketing informationrecommendation for every user.The software can be used for the recommendation application on Coloring RingBack Tone, web page, E-books, and other mobile phone service. The system employedthe Brower/Server structure and all the operating can be performed through the browser.The main results are as follows:1. We studied the purchase record and social behavior of cell phone users on theirpreference of interests in the song types, song styles, and singers.2. We proposed a social network analysis method on the diversity recommendationbased on the user tags behavior on Coloring Ring Back Tone download.3. Based on the research of the Hadoop platform to solve the problems of amassive users’ recommendation, and implemented our proposed social recommendationalgorithm in the distributed environment.4. We applied the system on the Coloring Ring Back Tone downloadrecommendation and obtained good results.The main goal of traditional recommender system is to improve the accuracy ofrecommendation, which leads to the tendency to recommend a small amount of popularitems. In order to enhance the User-System Viscosity, the diversity of recommendationlist for the old customers should be considered if the recommendation based onpopularity does not work well. Then it can meet the individual needs of customers. Therefore, how to simultaneously gain in both accuracy and diversity inrecommendation is the challenge that recommender system has to face. The maincontribution of this work is that we considered the social behavior of users in theirpurchase and proposed a method both accuracy and diversity in the recommendation.The research results can be used not only in the social recommendation, but also in thetraditional recommendation with long tail.