- 學位論文
考量讀者價值於圖書館書籍推薦：以協同過濾為基礎
Considering Reader Value In Library Book Recommendations: A Collaborative Filtering-Based Approach
摘要
推薦系統在電子商務領域扮演相當重要的角色，除可以協助銷售量的提昇，亦能夠針對顧客過去的偏好與交易行為，來提供個人化的服務，並增加顧客的滿意度。近年來，圖書館由圖書推薦系統的導入，來提高讀者使用圖書館的意願，以解決館藏使用率不高的問題，亦有相當不錯的成效。然而，在大學圖書館的環境下，讀者借閱動機不盡相同，例如課程中的作業或學期報告需要查找圖書館資料、或是於考試期間借閱參考書等，因此有些行為無法真正反映讀者的偏好。若未將讀者做篩選，而直接將所有讀者之借閱記錄投入進行圖書推薦系統之規則開發，將影響圖書推薦之品質，甚至產生無效的推薦。 為解決此問題，本研究首先透過圖書館借閱記錄找出忠實讀者群，運用行銷領域中衡量顧客價值的RFM模式，來進行讀者價值分析，並從中找出忠實讀者，接著再運用協同過濾的技術預測讀者-圖書借閱矩陣欄位之分數，本研究以MAE來衡量經過篩選的忠實讀者與未經篩選全部讀者之預測正確率的差異。
並列摘要
Recommended system in e-commerce plays a very important role, can help improve in sales, also able to provide personalized services from the past customers’ preferences and transactions, and increase customer satisfaction. In recent years, the library import recommendation system to increase the reader's willingness and solve the problem of library usage rate is not high, there are very good results. However, in the university library environment, readers borrowing motivation is different, such as course assignments or term papers need to find library materials, or borrow reference books during the examination period, so some of these acts can't really reflect the preferences of the reader .If the reader does not filtering, and direct use all reader's borrowing record into the recommended system of the book that to development the rules of books recommended system, will affect the quality of recommended books, and even ineffective recommendation. To solve this problem, this study first identified the loyal readers from the Library borrowing record, use RFM model of marketing field to measure customer value, and analysis readers value that to find loyal readers, then use collaborative filtering technology forecast column scores of readers - books borrowing matrix, this study use MAE to measure the forecasts accuracy differences in have filtered loyal readers and unfiltered reader.