隨著教育個性化需求的增加，如何根據學生的學習興趣和能力，為其推薦適合的
選修課程，一直是具有挑戰性的重要課題。本研究比較了傳統基於用戶的協同過濾方
法，大語言模型（Large Language Model, LLM），以及結合二者的混合模型在選修課
推薦中的應用效果。傳統協同過濾方法透過計算學生之間的皮爾森相關係數，基於相
似修課表現進行推薦。大語言模型則利用學生修課履歷和選修課列表，不僅參考成績
進行推薦，還能透過語義理解能力判斷待推薦課程是否與學生的修課背景相符。混合
模型進一步將協同過濾模型的推薦清單與學生的修課履歷結合作為提示輸入大語言模
型，生成新的推薦結果。實驗結果表明，混合模型在多項評估指標上均優於單一模
型，為未來的選課推薦系統設計提供了新的思路與改進方向。

With the increasing demand for personalized education, how to recommend suitable 
elective courses based on students' learning interests and abilities has become a challenging 
and important issue. This study compares the effectiveness of traditional user-based 
collaborative filtering (CF), a Large Language Model (LLM), and a hybrid model combining 
both approaches in the context of elective course recommendations. The traditional CF method 
recommends courses by calculating Pearson correlation coefficients between students and 
identifying those with similar course performance. The LLM, on the other hand, leverages 
students' course histories and elective course lists to provide recommendations. In addition to 
considering grades, the LLM can utilize semantic understanding to assess whether the 
suggested courses align with students' learning backgrounds. The hybrid model further 
integrates the recommendation list from the CF approach with students' course histories as 
prompts for the LLM, generating enhanced recommendations. Experimental results show that 
the hybrid model outperforms individual models across multiple evaluation metrics, offering 
new perspectives and directions for the design of future course recommendation systems.

目錄 
第一章 緒論 .................................................... 1 
1.1. 研究背景.............................................. 1 
1.2. 研究動機.............................................. 1 
1.3. 研究目的.............................................. 1 
1.4. 論文結構.............................................. 2 
第二章 文獻探討 ................................................ 3 
2.1. 推薦系統及技術 ......................................... 3 
2.2. 選課推薦系統 ........................................... 4 
第三章 研究方法 ................................................ 8 
3.1. 問題定義 ............................................... 8 
3.2. 推薦課程的方法 ......................................... 9 
3.2.1. 基於用戶的協同過濾模型推薦方法 .................... 9 
3.2.2. 大語言模型推薦方法 ............................... 11 
3.2.3. 混合模型推薦方法 ................................. 13 
3.3. 評估指標 .............................................. 15 
第四章 實驗設計與結果 ......................................... 19 
4.1. 資料集描述及後處理 .................................... 19 
4.2. 實驗環境 .............................................. 20 
4.3. 實驗結果............................................. 21 
4.3.1. 協同過濾模型參數調校設計 ........................ 21 
4.3.2. 協同過濾模型實驗 ................................ 22 
4.3.3. 大語言模型輸出檢查設計 .......................... 24 
4.3.4. 大語言模型實驗 .................................. 25 
4.3.5. 混合模型實驗 .................................... 27 
4.3.6. 三種模型的討論 .................................. 30 
4.3.7. 個案觀察 ........................................ 33 
第五章 結論 ................................................... 37 
5.1. 結語 .................................................. 37 
5.2. 管理意涵 .............................................. 38 
5.3. 研究限制 .............................................. 38 
5.4. 未來發展 .............................................. 38 
 參考文獻 
 附錄

圖目錄 
 圖1: 實驗流程圖…………………………………….………………………….8 
圖2: 大語言模型提示字串模板…………….…………….…………………..12 
圖3: 混合模型提示字串模板…………….…………….…………….……….14 
圖4: KNN以前一學期資料之預測結果….…………….…………….………23 
圖5: KNN以前二學期資料之預測結果….…………….…………….………23 
圖6: KNN以前三學期資料之預測結果….…………….…………….………23 
圖7: KNN以前四學期資料之預測結果….…………….…………….………23 
圖8: 大語言模型提示字串實際範例….…………….…………….………….25 
圖9: LLM以前一學期資料之預測結果….……………………………......…26 
圖10: LLM以前二學期資料之預測結果…………………………………….26 
圖11: LLM以前三學期資料之預測結果…………………………………….26 
圖12: LLM以前四學期資料之預測結果…………………………………….26 
圖13: 混合模型提示字串實際範例…………………………………………..28 
圖14: 混合模型以前一學期資料之預測結果………………………………..29 
圖15: 混合模型以前二學期資料之預測結果………………………………..29 
圖16: 混合模型以前三學期資料之預測結果………………………………..29 
圖17: 混合模型以前四學期資料之預測結果………………………………..29 
圖18: 資管系，三種模型的精確率-召回率曲線……………………………30 
圖19: 資工系，三種模型的精確率-召回率曲線……………………………30 
圖20: 財金系，三種模型的精確率-召回率曲線……………………………31 
圖21: 混合模型之MRR提升百分比 ………………………………………..32 
圖22: 混合模型之mAP提升百分比 ………………………………………..32 
圖23: 混合模型之NDCG提升百分比 …………………………………..33 

表目錄 
表1: 本實驗與其他相關推薦文獻比較 ............................................................ 7 
表2: 資料集欄位說明 ...................................................................................... 19 
表3: 各學系統計資料 ...................................................................................... 20 
表4: 協同過濾模型和KNN之參數K鄰居數評估結果 .............................. 22 
表5: KNN，不同資料量預測112-1學期之精確率 ....................................... 23 
表6: KNN，不同資料量預測112-1學期之召回率 ....................................... 24 
表7: LLM，不同資料量預測112-1學期之精確率 ....................................... 26 
表8: LLM，不同資料量預測112-1學期之召回率 ....................................... 27 
表9: 混合模型，不同資料量預測112-1學期之精確率 ............................... 29 
表10: 混合模型，不同資料量預測112-1學期之召回率 ............................. 30 
表11: 資管系，三種模型的MRR、mAP和NDCG實驗結果.................... 31 
表12: 資工系，三種模型的MRR、mAP和NDCG實驗結果.................... 32 
表13: 財金系，三種模型的MRR、mAP和NDCG實驗結果.................... 32 
表14: LLM 推薦效果最優之個案.................................................................... 34 
表15: KNN和混合模型推薦效果較優之個案 ............................................... 35 
表16: 混合模型推薦效果最優之個案 ............................................................ 36 

參考文獻 
Aggarwal, C. C. (2016). Recommender systems (Vol. 1). Springer.  
Alto, V. (2024). Building LLM Powered Applications: Create intelligent apps and 
agents with large language models. Packt Publishing Ltd.  
Behera, G., & Nain, N. (2022). Trade-off between memory and model-based 
collaborative filtering recommender system. Proceedings of the International 
Conference on Paradigms of Communication, Computing and Data Sciences: 
PCCDS 2021,  
Bobadilla, J., Ortega, F., Hernando, A., & Bernal, J. (2012). A collaborative filtering 
approach to mitigate the new user cold start problem. Knowledge-Based 
Systems, 26, 225-238.  
Brown, T. B. (2020). Language models are few-shot learners. arXiv preprint 
arXiv:2005.14165.  
Dwivedi, R., Anand, A., Johri, P., Banerji, A., & Gaur, N. (2020). Product based 
recommendation system on amazon data. Int J Creat Res Thoughts–IJCRT.  
Dwivedi, S., & Roshni, V. K. (2017). Recommender system for big data in education. 
2017 5th National Conference on E-Learning & E-Learning Technologies 
(ELELTECH),  
Ekin, S. (2023). Prompt engineering for ChatGPT: a quick guide to techniques, tips, 
and best practices. Authorea Preprints.  
Esteban, A., Zafra, A., & Romero, C. (2020). Helping university students to choose 
elective courses by using a hybrid multi-criteria recommendation system with 
genetic optimization. Knowledge-Based Systems, 194, 105385.  
Kolb, T. E., Wagne, A., Sertkan, M., & Neidhardt, J. (2023). Potentials of Combining 
Local Knowledge and LLMs for Recommender Systems. KaRS@ RecSys,  
Lops, P., De Gemmis, M., & Semeraro, G. (2011). Content-based recommender 
systems: State of the art and trends. Recommender systems handbook, 73-105.  
Millecamp, M., Htun, N. N., Jin, Y., & Verbert, K. (2018). Controlling spotify 
recommendations: effects of personal characteristics on music recommender 
user interfaces. Proceedings of the 26th Conference on user modeling, 
adaptation and personalization,  
Molina, L. E., & Bhulai, S. (2018). Recommendation system for netflix. Retrieved 
June, 6, 2021.  
Mondal, B., Patra, O., Mishra, S., & Patra, P. (2020). A course recommendation 
system based on grades. 2020 international conference on computer science, 
engineering and applications (ICCSEA),  
40 
Nguyen, V. A., Nguyen, H.-H., Nguyen, D.-L., & Le, M.-D. (2021). A course 
recommendation model for students based on learning outcome. Education 
and Information Technologies, 26, 5389-5415.  
Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., 
Rozière, B., Goyal, N., Hambro, E., & Azhar, F. (2023). Llama: Open and 
efficient foundation language models. arXiv preprint arXiv:2302.13971.  
Wu, L., Zheng, Z., Qiu, Z., Wang, H., Gu, H., Shen, T., Qin, C., Zhu, C., Zhu, H., & 
Liu, Q. (2024). A survey on large language models for recommendation. 
World Wide Web, 27(5), 60.  
Zhang, L., Luo, T., Zhang, F., & Wu, Y. (2018). A recommendation model based on 
deep neural network. IEEE Access, 6, 9454-9463.