利用協同式過濾模型建立考慮隱私的課程推薦系統

Collaborative Filtering Based Model for Privacy-Preserving Course Recommendation

李揚(Eric L. Lee)

指導教授 ： 林守德

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.02564

摘要

一個大學裡往往會有很多課程可供選擇，以台大為例，光2012年一年就有10572堂課可供選擇。對學生來說，在這些課程裡去做選擇是一件很花時間的事情。所以，這篇論文使用了學生過去的修課紀錄建立課程推薦系統。我們的課程推薦系統有兩個優點。第一點，跟之前的課程推薦系統的論文很不同的是，我們並沒有使用任何課程的資訊以及學生的成績或評價，而使有單純的使用學生選課的註冊紀錄，因此，保護了學生的隱私。第二點，跟之前的論文不一樣地方，ˊ之前的論文把任何一個物品當作是獨立的，但在我們這篇論文中，我們把每堂課當作不獨立的，所以更加提高了我們預測模型的表現。我們的實驗結果會顯示我們的課程推薦系統顯著地比傳統的推薦式系統還要來得好。

關鍵字

機器學習 ； 推薦系統 ； 人工智慧 ； 資料探勘 ； 教育

並列摘要

University students have to register for courses and usually there are many of those to choose from. It is time consuming for students check the course information for all courses before registration. As a result, this thesis proposes a recommender system to recommend courses to students based on the previous registration data of others. The advantage of our model is twofold. First, different from the previous works that require meta data about students or content information about courses, our model only needs the binary registration record of students for each course, thus protects the privacy of data provider. Second, different from the previous recommendation model that assumes items are independent, our model considers the courses-taken as a non-iid behavior to boost the performance. The experiment results show significant boost in our model comparing with the traditional recommender systems.

並列關鍵字

Machine Learning ； Recommender System ； Artificial Intelligence ； Data Mining ； Education

參考文獻

[1] Parameswaran, A., Venetis, P., & Garcia-Molina, H. (2011). Recommendation systems with complex constraints: A course recommendation perspective. ACM Transactions on Information Systems (TOIS), 29(4), 20.

[3] Ma, H., Zhou, D., Liu, C., Lyu, M. R., & King, I. (2011, February). Recommender systems with social regularization. In Proceedings of the fourth ACM international conference on Web search and data mining (pp. 287-296). ACM.

[5] Koren, Y., Bell, R., & Volinsky, C. (2009). Matrix factorization techniques for recommender systems. Computer, (8), 30-37.

[7] Bradley, A. P. (1997). The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern recognition, 30(7), 1145-1159.

[8] Pazzani, Michael J., and Daniel Billsus. "Content-based recommendation systems." The adaptive web. Springer Berlin Heidelberg, 2007. 325-341.