查詢結果分析
來源資料
頁籤選單縮合
題 名 | 利用資料探勘技術發掘圖書館個人化之書籍推薦=Using Data Mining Techniques to Discover Personalized Book Recommendation for Library |
---|---|
作 者 | 陳垂呈; | 書刊名 | 教育資料與圖書館學 |
卷 期 | 43:1 民94.09 |
頁 次 | 頁87-107 |
分類號 | 023.6 |
關鍵詞 | 資料探勘; 分類分析; 借閱資料; 書籍推薦; Data mining; Classification analysis; Borrowing history records; Book recommendations; |
語 文 | 中文(Chinese) |
中文摘要 | 本論文以讀者之借閱資料為探勘的資料來源,每一筆借閱資料包含讀者曾借閱過的書籍與其興趣度,並以某一讀者之借閱資料為分析的目標,利用資料探勘(data mining)技術中的分類分析(classification),探討如何發掘此一讀者個人化的書籍推薦。在探勘過程中,比對此一讀者借閱資料與其他借閱資料的相似度,依據其是否符合所設定的條件,來分別設定其與此一讀者借閱資料的關聯性為「高」或「低」,並視其他非此一讀者曾借閱過的書籍項目為影響屬性,然後對讀者的借閱資料進行分類分析。首先,只考量讀者曾借閱過的書籍項目,然後針對借閱資料進行分類分析,藉由所建立的決策樹(decision trees),可得知那些屬性與此一讀者之關聯性為高,藉以發掘出此一讀者個人化最適性的書籍推薦。再者,考量讀者對曾借閱過之書籍的興趣度,分別將每一書籍分解成其興趣度之數量的項目屬性,然後視分解後非此一讀者曾借閱過之項目屬性為欲分類的屬性,並進行分類分析,藉由所建立的決策樹,可得知那些分解後之項目屬性與此一讀者的關聯性為高,藉以發掘出包含有興趣度之此一讀者最適性的書籍推薦。此探勘結果,對圖書館在擬訂最適性之讀者個人化書籍推薦時,可以提供非常有用的參考資訊。 |
英文摘要 | In this paper, we use readers borrowing history records as the source data of mining. Each borrowing history record contains a reader ever borrowed books with the degree of interest. We let one reader as the target of mining and use classification analysis to discover the personalized book recommendations for the reader. In the mining process, we compute the degree of similarity of borrowing history records between the reader and other. If the degree conform the given condition, we assign the association level between the both readers is “high”. Otherwise, it is “low”. For books not borrowed by the reader, we treat those books as attributes for classification. First, we only consider readers ever borrowed books, and classify the borrowing history records to construct a decision tree. We can find the association level to be “high” between some attributes and the reader according to the decision tree. It is the basis to discover the most adaptive book recommendations for the reader. Moreover, we consider books with readers interests in the borrowing history records. Each book is divided to u unit items where u is the degree of the interest, u is positive integer, and the degrees of interest of these items are, respectively, from 1 to u. For books not borrowed by the reader, we divide those books to unit items and treat those items as attributes for classification. We can construct a decision tree after classifying the borrowing history records. According to the decision tree, we can find the association level to be “high” between some attributes and the reader. It is the basis to discover the most adaptive book recommendations for considering the reader's interesting. The results of the mining can provide very useful information to recommend the most adaptive books for individual reader. |
本系統中英文摘要資訊取自各篇刊載內容。