Personalized Recommendation Algorithm Based On Data Mining Technology In The Digital Libray

With the development of Internet, the information resources of Digital Library currently become richer. But with the expansion of information, the phenomenon of"information overload"and"mislead"were appeared. In order to provide the characteristics of their preference information to users, personalized recommendation technology becomes an important research field in digital library and attracts many researchers'attention at the same time. Collaborative filtering technology and association rules recommendation technology are the key research point in personalized recommendation field Simultaneously, the research for these issues are focused in this paper.Firstly, the existing problems of the recommended collaborative filtering technology, were analyzed. With the increasing number of digital library users and book resources, the evaluate matrix was extremely sparse in the whole project since the differences between users'professional background and recommendations made more errors with user's interest, A recommendation algorithm based on clustering technology is proposed in this paper. This algorithm aggregates the library resources by combining the K-means technology and the self-organizing map technology. It reduces the scope of neighbor searching and the number of library resources which need forecast. The problems of sparsity were successfully resolved and the accuracy of the recommendation was improved.Secondly, in order to enhance the rate of book borrowing, association rules were used to obtain users'hobbies. The analysis of inhomogeneous readers'preference provides related books to users. If the FP-growth was directly used, the huge amounts of substantial association rules were made. It would place an heavy burden to recommended system according to these rules. At the same time,the rules may cause substantial duplication of recommendation and make a lot of redundancy terms. A satisfactory effect on the recommendation was achieved through improving the FP_growth algorithms and combining online and offline recommended methodologies.Finally, it also gives out the analysis and verification to all the technology which were mentioned in the paper. Subsequently, the prospects for future research were made.