Research On Personalized Recommendation Algorithm Based On Rough Set Theory

With the development of the computer technology and network, we can't do any thing without network in our life, work and study, so the personalized recommendation system arises quickly.It can analyze users'registered information and web-browsing behaviors to judge users'possible needs, and then do the personalized recommendation to users in some personalization recommend ways.The application of the system can simplify the webs'service and develop the users'loyalty.We mainly focus on personalized recommendation algorithms in this paper.After analyzing and comparing the existing algorithms, we design a rule-acquisition algorithm based on binary-distinguish matrixes, against incompletion and redundancy in the rule derived in personalized recommendation based on association rule, and we test its feasibility with Movielens database.The thesis mainly includes three aspects as follows:Firstly, we learn the status quo and trend of theoretical research about personalized recommendation after reading a great number of literatures at home and abroad, introduce and compare some common personalized recommendation algorithm in the thesis, and mainly study the personalized recommendation based on association rules, whose core is to get minimal rule.Secondly, we study the rough set theory which can deal with a large number of uncertain data about association rules.We describe the developing process, basic concepts, common methods, application software and theoretical characters about rough set theory, mainly focusing on some attribute-reduction and rule-acquisition algorithm in rough set theory.Based on binary-distinguish matrixes, we propose an improved rule-acquisition algorithm.Thirdly, in order to test the new algorithm's feasibility, we design a simulation experiment with Movielens database to compare the results from the new algorithm, distinguish matrixes and existing methods of ROSETTA.