推薦關係在許多商業中起著重要的作用，在本文中，我們提出的模型將
線性混合效應模型作為推薦系統的基礎模型。在這個條件下所提出的模型便於計算，不但對於數量大的數據集也是可行的，而且對於冷啟動問題相對應的新用戶或者新項目的情形都是實用的。為了驗證我所提出的模型，我選用了MovieLens作為數據集。在這個數據集中包括了由943位用戶和1,682部電影，一共100,000個評分。數據集中還提供了有關用戶和電影的信息，例如用戶的年齡、性別；電影的類別等等。通過採用這些特征徵信息來解決冷啟動的問題。在計算中，由於數據過多，使用貝葉斯模型或者EM演算法的傳統推薦系統雖然會具有更加準確的結果，但是計算負擔過大。所以我們提出的線性混合模型是可以用簡單有效的方法去解決問題的。

Recommender (or recommendation) system plays a critical role in many business models. In this this thesis, we consider linear mixed effect model as a base model for recommender system. Under this framework, the proposed model is easy to compute, feasible for large data, and practical for new user/item scenario which corresponds to the so-called cold start problem. To demonstrate the proposed model, we consider MovieLens database consisting 100,000 ratings, rated by 943 users among 1682 movies. Features about users and movies are also available in this database. The cold start problem is conquered by adopting the information of these features. Due to the excessive data, conventional recommender systems using Bayesian models or EM algorithms are impractical although they may have more accurate results. The proposed method can be a simple and efficient base of recommender systems.

[1] Srebro, N., Rennie, J. and Jaakkola, T. (2004). “Maximum margin matrix factorizations”, Proceedings of the 17th International Conference on Neural Information Processing Systems, pp. 1329-1336.
[2] Srebro, N.and Rennie, J. (2005), “Fast maximum margin matrix factorization for collaborative prediction”, Appearing in Proceedings of the 22 nd International Conference on Machine Learning, Bonn, Germany, pp. 713-719.
[3] Agarwal, D. and Chen, B. C. (2006), “Statistical Methods for Recommender Systems (1 ed.)”, Cambridge University Press, pp. 120-141.
[4] Bell, R., Koren, Y. and Volinsky, C. (2007), “Modeling relationships at multiple scales to improve accuracy of large recommender systems”, Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining, pp. 95-104.
[5] Park, S. T., Pennock, D., Madini, O., Good, N. and DeCoste, D. (2006), “Naïve filter bots for robust cold-start recommendations”, Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining, pp. 699-705.
[6] Agarwal, D. and Merugu, S. (2007), “Predictive discrete latent factor models”, Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining, pp. 26-35.
[7] Zhang, Y. and Koren, J. (2007), “Efficient Bayesian hierarchical user modeling for recommendation system”, Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval. pp. 47-54.
[8] Agarwal, D. Chen, B. C., Elango, P., Motgi, N., Park, S. T., Ramakrishnan, R., Roy, S. and Zachariah, J. (2008), “Online models for content optimization”,published at the Neural Information Processing Systems Conference. pp. 17-
24.
[9] Ruslan, S. and Andriy, M. (2008), “Probabilistic matrix factorization”, Proceedings of the 20th International Conference on Neural Information Processing Systems, pp. 1257-1264.
[10] Salakhutdinov, R. and Mnih, A. (2008), “Bayesian probabilistic matrix factorization using Markov chain Monte Carlo”, Proceedings of the 25th international conference on Machine learning, pp. 880-887.
[11] Agarwal, D. and Chen, B. C. (2009), “Regression-based latent factor models”, Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining, pp. 19-28.
[12] Stern, D., Herbrich, R. and Graepel, T. (2009), “Matchbox: Large Scale Online Bayesian Recommendations”, Proceedings of the 18th International Conference on World Wide Web, pp. 111-120.
[13] Zhang, L. Agarwal, D. and Chen, B. C. (2011), “Generalizing Matrix Factorization Through Flexible Regression Priors”, Proceedings of the fifth ACM conference on Recommender systems, pp. 13-20.