改良式項目為基礎的協同過濾推薦系統

An Improved Item-based Collaborative Filtering Recommendation System

黃健倫(Jian-Lun Huang)

指導教授 ： 林榮禾

國立臺北科技大學/管理學院/工業工程與管理研究所/碩士(2009年)

摘要

隨著網際網路的盛行及資訊的數位化，網路上充斥著各式各樣的資訊，使用者必須花很多時間去找他們需要、感興趣的資訊，造成資訊超載的問題。因此利用推薦系統依據使用者的興趣、喜好推薦使用者感興趣的資訊、商品，提供使用者更適切的服務以提升顧客忠誠度成為一個重要的研究課題。
在推薦系統研究中，常用的方法包括內容為基礎式推薦系統、使用者為基礎的協同過濾式推薦系統、項目為基礎的協同過濾推薦系統、混合式推薦系統，其中項目為基礎的協同過濾推薦系統是目前發展的最成功的推薦系統且廣泛應用於業界，例如亞馬遜，然而此方法有兩個盲點：(1)忽略目標使用者與用來計算項目相似度的使用者的興趣、喜好的差異。(2) 忽略個人興趣會隨著時間而改變。因此本研究提出結合使用者相似度與時間權重於項目為基礎的協同過濾推薦方法(Hybrid User-Weight and Time-Weight Item-Based Collaborative Filtering, HUTCF)改良上述兩項盲點。實驗結果顯示，本研究所提出的方法準確率高於項目為基礎協同過濾推薦方法且隨著使用者使用系統的時間增加，提出的方法推薦效能也越好，因此相信在實務應用上，透過本究所提出的方法來建構推薦系統，能夠準確的推薦使用者感興趣的資訊。

關鍵字

推薦系統 ； 資料探勘 ； 協同過濾

並列摘要

With the advance in information technology and the emergence of the Internet, users can share information with each other online. Therefore, users must spend a lot of time to find their needs and information that users interested, resulting in the problem of information overload. Recommendation system is a tool to alleviate the problem of information overload, It recommend users the potential demand and help users filtering information, quickly find the information that their needs.
The common method in recommendation system includes content-based collaborative filtering, user-based collaborative filtering, item-based collaborative filtering, hybrid approach recommendation system. Item-based collaborative filtering recommendation system is currently the most successful recommendation system and it is widely applied in business, ex:Amazon. The method, however, fails to take into consideration of two important issues: the differences between users in their interests and hobby as well as the changes of users’ interest over time. Therefore this study combined user similarity weight and time weight into item-based collaborative filtering recommendation method to improve the two blind spots. The result of the experiment show that the method proposed in this study outperforms item-based collaborative filtering recommendation method. Therefore, we believe that using this method to construct the recommendation system, it can recommend accurately information to users.

並列關鍵字

Recommendation System ； Data Mining ； Collaborative Filtering

參考文獻

[2] 周寬怡，以內容分析法獲取推薦系統中使用者profile之研究，碩士論文，國立成功大學資訊管理研究所，台南， 2003。

[3] 張毓倫，個人化顯隱性知識推薦方法之研究，碩士論文，國立成功大學資訊管理研究所，台南， 2003。

[5] A. Albadvi and M. Shahbazi, “A hybrid recommendation technique based on product category attributes,” Expert Systems with Applications, vol.36, no.9 pp.11480-11488.

[6] A. Mild and T. Reutterer, “Collaborative filtering methods for binary market basket data analysis,” Lecture Notes in Computer Science, 2001, pp.302–313.

[7] B. Krulwich and C. Burkey, “The InfoFinder agent: Learning user interests through heuristic phrase extraction,” IEEE Intelligent Systems Journal, vol.12, no.5, 1997, pp.22-27.