Research And Design Of Recommendation System

With the rapid development of the Internet industry.Every day people has to face a huge amount of data. How to find useful data becomes a problem that needs to be solved. Recommendation system uses the theory of data mining to explore the useful information and these data will be pushed to the users by a list.The main work of this paper was as follows.(1) At first,this paper introduces the background of recommendation system and some results of previous work. Because of factorization recommendation algorithm has the problem of accuracy.We put forward the NFDRA,for improving the effort of factorization algorithm.In research of recommendation system the factorization recommendation algorithm has been successfully applied, which has been applied in the Amazon and other e-commerce sites. But most of the factorization algorithms do not consider time effects. In this paper we summarized the factors of time effect and propose a new algorithm named NFDRA. Compared with other factorization recommendation algorithms,the NFDRA has its own advantages on the metrics of RMSE,precision and recall.(2) Then,in order to improve the sorting effects of factorization algorithm,we put forward the Pairwise-SVD algorithm.This algorithm combined the theory of learning to rank.Although factorization algorithms could predict the score of users, but it does not consider how to generate recommendation list. Here we put forward the Pairwise-SVD recommendation algorithm which based on the theory of learning to rank. And through the comparison of three other algorithms on metrics of MRR and Kendall-tau, we can make a conclusion that the Pairwise-SVD recommendation algorithm has a better scheduling effect.(3) At last,we designed a movie recommendation system by using the theory of software engineering.Although many researchers have done the work of algorithms,but few of them care about how to design and realize a true recommendation system. So in this paper, by using the theory of software engineering we designed a movie recommendation system.Finally, we summarize the work and point out the outlooks and requirements for future work.