圖書館借閱記錄探勘系統

Abstract

隨著網際網路的發展與電腦科技的日益進步，資訊數位化已成為世界的趨勢，電子圖書館也在這股資訊潮流下日漸成熟，而如何利用電腦技術以提昇圖書館對讀者的服務品質亦成為各圖書館努力的目標。 由於圖書館的借閱記錄有如讀者使用圖書館資源的最佳證據，因此本論文藉由分析交通大學圖書館的借閱記錄以了解讀者借閱館藏的關聯性，再根據以往讀者借閱的關聯性將館藏有效地推薦給其他讀者，讓交通大學圖書館在讀者探索知識的過程中扮演著積極主動的角色。 本研究根據圖書館借閱記錄的特性，選擇適合圖書館的相關規則演算法並加以改良應用至廣義相關規則探勘(Generalized Association Rule Mining)及多重最小支持度廣義相關規則探勘(Generalized Association Rule Mining with Multiple Minimum Supports)，實作適合圖書館的資料探勘系統「圖書館借閱記錄探勘系統」。讓館員藉由輸入讀者借閱記錄得到最新的館藏借閱相關規則，針對不同系所的讀者找出不同的相關規則。亦應用「中國圖書分類法」找出讀者借閱關聯類別，且可針對不同階層的類別設定不同的最小支持度門檻值，探勘多重最小支持度廣義相關規則，並結合交通大學個人化數位圖書資訊環境 (PIE@NCTU) 將相關館藏推薦給讀者。

With the rapid development of Internet, digitization has been a world trend. The proliferation of Internet also encourages the development of electronic libraries. In the era of new information technology, how to make use of computer technology to provide readers better services has been the target of all libraries. The borrowing history of patrons is one excellent evidence to track patrons’ interests, in view of this, we aim at finding the association of the collections in National Chiao Tung University (NCTU) Library by analyzing the borrowing history records of NCTU Library. Furthermore, we recommend the associated collections to patrons according to the findings. We expect that NCTU Library can play an active role in the knowledge discovery of NCTU patrons. In order to achieve the above goal, this thesis chooses the suitable association rule algorithm H-Mine for mining library records and modifies H-Mine to generalized association rule mining and association rule mining with multiple minimum supports. We also implement a data mining system suitable for libraries, the Library Borrowing History Records Mining System. Librarians can get the latest association rules by inserting new library borrowing history records into database, and find different association rules according to patrons of different departments and institutes. This system also utilizes “New Classification Scheme for Chinese Libraries” to mine associated categories and collections. Furthermore, this system integrates the association rules into a persobalized system, PIE@NCTU (Personalized Information Environment for National Chiao Tung University Library), to recommend associated collections to patrons.