考慮分類號結構為主的圖書借閱推薦方法

機構典藏 > College of Business and Management > Graduate Institute & Department of Information Management > Thesis > Item 987654321/105550

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/105550

Title:	考慮分類號結構為主的圖書借閱推薦方法
Other Titles:	Subject-code-based book recommendation method
Authors:	鄭皓澤;Tey, Haw-Zhe
Contributors:	淡江大學資訊管理學系碩士班魏世杰;Wei, Shih-Chieh
Keywords:	圖書借閱推薦;圖書分類號;協同推薦;分類號相似度;Book Borrowing Recommendation;Library Catalogue Subject Code;Collaborative Recommendation;Subject Code Similarity
Date:	2015
Issue Date:	2016-01-22 14:58:54 (UTC+8)
Abstract:	分類表是具體而微的人類知識表徵，其類目設計與結構的一致性具有一定的公信力，但透過探勘借閱行為發現許多相關書籍並沒有歸在同一分類號層級中。由於圖書分類號代表書籍的知識領域類別，故借閱者借閱某書可視為其對某分類號代表的知識領域有偏好。傳統協同推薦中，所推薦出的清單可能含有用戶未偏好的書籍。基於上述原因，本文在過去研究的人推薦物內嵌物推薦物架構上，分別加入圖書分類號協同相似度、圖書分類號固有相似度、同儕熱門度之考量，以觀察其對提升圖書推薦系統的效果。結果發現圖書分類號協同相似度搭配圖書分類號固有相似度在意外性表現最好，Top-N精確率以圖書分類號協同相似度、圖書分類號固有相似度、同儕熱門度的組合最顯著，整體精確率則以圖書分類號依字典排序距離、圖書分類號固有相似度、同儕熱門度的組合最佳。 The library catalogue classification system is a characterization of human knowledge which contains a consistent and credible design of hierarchical subject codes. But through mining of borrowing behavior, it is found that many relevant books do not belong to the same subject code. As each subject classification code represents a certain category of knowledge domain, when a user borrows a book, it means that he has preference for the book’s knowledge domain. In traditional collaborative recommendation, the recommended list may contain books the user dislikes. Due to the above reasons, based on a past framework which allows item-to-item recommendation embedded in user-to-item recommendation, this work considers subject code collaborative similarity, subject code native similarity, and peer popularity to improve the recommendation. Our experimental results show that in terms of unexpectedness, recommendation using subject code collaborative similarity and subject code native similarity performs the best. In terms of top-n precision, recommendation using subject code collaborative similarity, subject code native similarity and peer popularity performs the best. In terms of average precision, recommendation using subject code lexicographic distance, subject code native similarity and peer popularity performs the best.
Appears in Collections:	[Graduate Institute & Department of Information Management] Thesis

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	94	View/Open

Loading...