The Research And Achievement Of College Library New Book Recommendation System

College library is an important learning and research environment and also an importantchannel that teachers and students can access reference resources. Now most of college librarieshave been established at all levels of web-based information service system, but most of thelibrary service systems just provide some simple inquiry functions, such as looking up the libraryinformation and readers’ loan information, or judge which kind of book will attract a reader sinterest long based on his loan record, or recommend related information of the books with thesame kind simply when a reader borrows books. Nevertheless, these are not effective to providepersonalized information service aimed at readers preferences and specific demand. When wecollect some certain kind of new books, which group shall we inform and what characteristics dothey own? What s more, after we inform them, will we be confident that they will come toborrow these books? Thus, we expect to solve this problem.The main content of this article can be divided into4parts:Part one is literature review, mainly putting forth the knowledge discoveries and theoreticalbasis of data mining techniques.Part two is to dig up the framing method of data mining process, describing each part ofdata mining in readers’ loan record of college libraries.Part three is algorithm discussion. Through the comparison in the complexity andeffectiveness, the most appropriate rules analysis approach can be found in each search algorithm,such as SBSP, SBMP, LatSBMP and MBMP. We try to find out the following rules: attribute oneof the readers, attribute….Attribute N=>bibliography support=X%confidence=Y%. So that,as for a new book, we can classify this book into a certain bibliography and then recommend thisbook to the related readers group based on the above rules derived by the same kind ofbibliography with this book. Part four is achievement and effect evaluation for new book recommendation system andthe outlook for the future work.This thesis is innovative for recommendation rules. We find out the reader s rulescorresponding to the book category based on one book or book hierarchy level instead of thereaders of prior classification...