Abstract: This paper used data mining technology to analyze the readers’ historical records in the library at a timeaccording to the borrowing data of a certain university library. Through the statistical analysis of records at a time, it inferred the readers' behavior characteristics of different level, and determined their different demands of libraryliterature information service according to the behavior characteristics. Based on the demand information, library could provide real-time personalized professional service, as well as scientific basis for the optimization of library collection, so as to improve the literature resource guarantee rate of university library..

Key words: Reader group