The Research Of User Interest Model Based Recommendation Algorithm And System Realization

With the development of social network based on Internet, it products largeamount of information, information overload appears in social network, it’s difficult forusers to find out contents that they are interested in, so it’s necessary to applypersonalized recommendation to social network.Now researches of social recommendation mainly use social information toconstruct user interest model, and recommend according to user’s interest. User’sinterest is one of main factors in personalized recommendation, so constructing user’sinterest precisely is important to personalized recommendation. This paper analyzedexisting algorithms of social recommendation deeply, and proposed a kind ofrecommending model of microblogging’s contents based on interest similaritycomputation. This model includes classifying user interest’s types and measuring users’tie-strength, and utilize user’s own interest and social behavior comprehensively tocompute interest, so as to improve the accuracy of recommending.The paper proposes a classification calculation mode of constructing user’s interestmodel dynamically according to user’s interest dynamical changing, and defines user’sinterest by long-term interest and short-term interest. Tie-strength between users is animportant character of social relationship, so when recommending in social network,we thought tie-strength as an influence factor. According to characteristic of socialnetwork such as weibo, users’ relationship is based on common interest, we use interestsimilarity to replace users’ tie-strength. Combined with the microblogging application,this paper proposed a method of using friends’ interest to expand user’s interest forimproving precision and stability of recommending of low activities users.When to realize the recommending system, to improve system resource utilization,we designed offline computation module according to user regular rest that they rarelyuse system at middle night. The offline computation module computes user’s interestmodule dynamically at middle night, which realizes optimized schema of incrementalupdating offline, can improve system resource utilization and system response speed.This paper developed distributed recommendation system based on weibo’scontents recommendation algorithm, and test effects of algorithm and system online.We used cache and offline computation to improve system response speed, anddeveloped related function modules. According to results of tests online, using friends’interest expanding users’ interest and interest similarity replacing users’ tie-strength allimprove the precision and stability of recommendation. After testing, when dealingwith high concurrent requests, the performance of distributed recommending system ofthis paper improves obviously.