近年來隨著巨量資料的蓬勃發展以及資訊科技快速的進步，許多企業網路
或電子商務店家，會根據消費者購買的紀錄、偏好來推薦消費者適合的產品，因此如何在眾多的項目中找到自己需要的商品就顯得格外重要，推薦系統就是在這個環境中逐漸發展成型。推薦系統(Personalized Recommender Systems)主要分成兩大類，第一類是基於內容的推薦系統(Content-based Recommendation)，為系統基於用戶的評價、個人資料去學習特定用戶的偏好， 並且推薦相似度高的商品給其用戶，當用戶的瀏覽紀錄越多其推薦效果越佳。第二類是協同過濾推薦系(Collaborative Filtering Recommendation)，為系統利用用戶彼此之間或是項目與項目之間的關聯性來推薦商品，其概念為若用戶喜歡的東西類似，則會推薦特定用戶類似的東西，此方法的優點為不需要用戶的評分紀錄即可運作而
第三類方法為混合式推薦系統，此方法主要利用矩陣分解結合模型文字提取模型本論文主要採用第三類混合式推薦系統，能更加準確的預測用戶對商品的評分。
因此本論文提出階層式深度學習網路結合機率矩陣分解法的模型，
首先我們先擷取出用戶評論的資料集，輸入到我們建構的深度學習網路架
構，經由Attention機制給予每一字詞給予權重分配，關注每篇評論的字詞權重，最後與矩陣分解法的潛在隱藏因子向量做交互訓練。RMSE為論文的實驗指標，效果明顯優於第四章的各大方法，本論文將取五個隨機種子產生器的平均值當作實驗效果指標，可得知我們的效能比Probabilistic Matrix Factorization好8%，比Convolutional Matrix Factorization for Document Context-Aware
Recommendation好5%，比Dual-Regularized Matrix Factorization with Deep Neural Networks Item好4%，因此更能精準預測使用者對商品的評分。

Before the recommender system generating recommendation results, we need to filter out useful information from the big data, but the information may be explicitor implicit. The results of the traditional Matrix Factorization, usually consider the accuracy of the first few results. Therefore, considering the global covariance is often inaccurate. Studies have shown that the probabilistic matrix factorization can produce good recommendations, but the disadvantage is that it only considered the
product rating in the model. And in other studies, the recommended results using Probabilistic Matrix Factorization are not well explained. Exampleblei et al. (2003) proposed the Latent Dirichlet Allocation to generate the topic model, which always ignores the context of each word. Therefore, Wang and Blei (2011) proposed the
Collaborative Topic Modeling for recommending scientific articles, which combining the Probabilistic Matrix Factorization model and the Latent Dirichlet Allocation
model to generate the user recommendation. The effect is significantly better than the original Matrix Factorization method and has more accurate predictions.
To clarify the user’s preferences, the methods related to sentimental text analysis can perform well and also can extract the product feature and sentimental text from the comment. The natural language process has been extensively researched since the millennium. The deep learning network has excellent performance in text mining and topic analysis. Most of the text sentiment analysis methods can’t give the order of the best or bad in the feature of the user’s favorite products. Therefore, this thesis combines deep learning sentiment analysis and Probabilistic Matrix Factorization to find out the user’s true preferences.

Bahdanau, D., Cho, K., and Bengio, Y. (2014). Neural machine translation by jointly
learning to align and translate. arXiv e-prints, abs/1409.0473.
Bouvrie, J. (2006). Notes on convolutional neural networks.
Cho, K., van Merri¨enboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk,
H., and Bengio, Y. (2014a). Learning phrase representations using RNN encoder–
decoder for statistical machine translation. In Proceedings of the 2014 Conference
on Empirical Methods in Natural Language Processing (EMNLP), pages 1724–1734,
Doha, Qatar. Association for Computational Linguistics.
Cho, K., van Merri¨enboer, B., G¨ulc¸ehre, C¸ ., Bahdanau, D., Bougares, F., Schwenk,
H., and Bengio, Y. (2014b). Learning phrase representations using rnn encoder–
decoder for statistical machine translation. In Proceedings of the 2014 Conference
on Empirical Methods in Natural Language Processing (EMNLP), pages 1724–1734,
Doha, Qatar. Association for Computational Linguistics.
Dong, X., Yu, L., Wu, Z., Sun, Y., Yuan, L., and Zhang, F. (2017). A hybrid collaborative
filtering model with deep structure for recommender systems. In AAAI.
Kim, D., Park, C., Oh, J., Lee, S., and Yu, H. (2016). Convolutional matrix factorization
for document context-aware recommendation. In Proceedings of the 10th ACM
Conference on Recommender Systems, RecSys ’16, pages 233–240, New York, NY,
The USA. ACM.
Koren, Y., Bell, R., and Volinsky, C. (2009). Matrix factorization techniques for recommender
systems. Computer, 42(8):30–37.
Liang, D., Krishnan, R. G., Hoffman, M. D., and Jebara, T. (2018). Variational autoencoders
for collaborative filtering. In Proceedings of the 2018 World Wide Web
The conference, WWW ’18, pages 689–698, Republic and Canton of Geneva, Switzerland.
International World Wide Web Conferences Steering Committee.
McAuley, J. and Leskovec, J. (2013). Hidden factors and hidden topics: understanding
rating dimensions with review text. In Proceedings of the 7th ACM Conference on
Recommender Systems, RecSys ’13, pages 165–172, New York, NY, USA. ACM.
Ning, X. and af, L. Y. (2018). Rating prediction via generative convolutional neural
networks based regression. Pattern Recognition Letters.
Qian, N. (1999). On the momentum term in gradient descent learning algorithms. Neural
Networks, 12(1):145–151.
Resnick, P., Iacovou, N., Suchak, M., Bergstrom, P., and Riedl, J. (1994). Groupon's:
Open architecture for collaborative filtering of netnews. In Proceedings of the
1994 ACM Conference on Computer Supported cooperative work, CSCW ’94, pages
175–186, New York, NY, USA. ACM.
Salakhutdinov, R. and Mnih, A. (2007). Probabilistic matrix factorization. In Proceedings
of the 20th International Conference on Neural Information Processing Systems,
NIPS’07 pages 1257–1264, USA. Curran Associates Inc.
Sarwar, B., Karypis, G., Konstan, J., and Riedl, J. (2001). Item-based collaborative
filtering recommendation algorithms. In Proceedings of the 10th International Conference
on World Wide Web, WWW’01 pages 285–295, New York, NY, USA. ACM.
Schafer, J. B., Frankowski, D., Herlocker, J., and Sen, S. (2007). Collaborative Filtering
Recommender Systems, pages 291–324. Springer Berlin Heidelberg, Berlin,
Heidelberg.
Sukhbaatar, S., Szlam, A., Weston, J., and Fergus, R. (2015). Weakly supervised memory
networks. CoRR, abs/1503.08895.
Tang, D., Qin, B., and Liu, T. (2016). Aspect level sentiment classification with deep
memory network. In Proceedings of the 2016 Conference on Empirical Methods in
Natural Language Processing, pages 214–224, Austin, Texas. Association for Computational
Linguistics.
Wang, C. and Blei, D. M. (2011). Collaborative topic modeling for recommending scientific
articles. In Proceedings of the 17th ACM SIGKDD International Conference
on Knowledge Discovery and Data Mining, KDD ’11, pages 448–456, New York,
NY, USA. ACM.
Wang, H., Shi, X., and Yeung, D. (2016a). Collaborative recurrent autoencoder: Recommend
while learning to fill in the blanks. CoRR, abs/1611.00454.
Wang, H., Wang, N., and Yeung, D. (2014). Collaborative deep learning for recommender
systems. CoRR, abs/1409.2944.
Wang, Y., Huang, M., Zhu, X., and Zhao, L. (2016b). Attention-based lstm for aspectlevel
sentiment classification. In Proceedings of the 2016 Conference on Empirical
Methods in Natural Language Processing, pages 606–615, Austin, Texas. Association
for Computational Linguistics.
Wu, C., Wang, J., Liu, J., and Liu, W. (2016). Recurrent neural network-based recommendation
for time heterogeneous feedback. Knowledge-Based Systems, 109:90 –
103.
Wu, H., Zhang, Z., Yue, K., Zhang, B., He, J., and Sun, L. (2018). Dual-regularized matrix
factorization with deep neural networks for recommender systems. Knowledge-
Based Systems, 145:46 – 58.
Yang, Z., Yang, D., Dyer, C., He, X., Smola, A. J., and Hovy, E. H. (2016). Hierarchical
attention networks for document classification. In HLT-NAACL.
Zhang, L., Luo, T., Zhang, F., and Wu, Y. (2018). A recommendation model based on
deep neural network. IEEE Access, 6:9454–9463.