圖書館個人化館藏推薦系統

Abstract

推薦系統是網路商店上常運用的技術，主要用來提高顧客的購買慾望。我們將此技術應用在圖書館上，希望能藉由個人化館藏推薦系統推薦給讀者圖書館中其有興趣的館藏，幫助讀者使用圖書館資源。同時也希望圖書館這個新的館藏推薦服務能增加館藏的利用率，並提高圖書館的價值。 在本論文中，將利用資料探勘 (Data Mining) 中關聯規則探勘 (Association Mining) 的技術，從讀者的借閱歷史檔中找出頻繁項目集 (Frequent Itemsets) 和關聯規則 (Association Rules)。從這些探勘的結果中，分析得知讀者與讀者間和館藏與館藏間存在的關係，由此取得讀者的興趣。接著先利用推薦系統中常用的協力式過濾 (Collaborative Filtering) 找出給讀者的推薦書目清單，再以內容導向過濾 (Content-Based Filtering) 的方法將推薦清單依照讀者興趣做個人化的排序，最後能找出合適的館藏推薦給讀者。 個人化館藏推薦系統目前實作於交通大學浩然圖書館的個人化資訊環境myLibrary中(http://mylibrary.e-lib.nctu.edu.tw/)。並且從收集回來的讀者回應中發現，這套系統的確可以有效地依照讀者的興趣將館藏推薦給讀者。

Recommender systems are popularly being used in e-Commence to encourage consumers to purchase more products. In library, recommender systems can also be used to help patrons find collections to borrow, and increase the value of library. In this paper, we propose a personalized recommender system for library, which combines collaborative filtering and content-based filtering recommendation methods. We hope patrons can find books which fit their preferences through this personalized recommender system. In this recommender system, first, we use association mining to discover association rules between users and between books. Second, we take these associations rules to find user groups which have the same preference for collaborative filtering, and build representation model for content-based filtering. Finally, we use collaborative filtering to find a recommendations list which contains books collecting from other patrons’ borrowing history, and we would re-rank this recommendations list by content-based filtering to cater to user’s preference. We implemented this recommender system on National Chiao Tung University (NCTU) library’s personalized information environment “myLibrary” (http://mylibrary.e-lib.nctu.edu.tw/), and the feedback received from users of library shows that most users satisfied with our recommendations.