社群網路的興起，許多的消費者樂於在社群媒體上討論分享，表達自己對產品的意見。企業可透過大量的網路評論分析市場上消費者對產品各項特徵的喜好與優缺點，但在過去的文獻中較少應用深度學習於中文評論的情感分析上。
    本論文的貢獻為透過文本分析建構出專屬於智慧型手環領域的情感意見詞典，並利用深度學習遞迴神經網路長短期記憶技術於智慧型手環口碑情感分析，與貝氏演算法、支援向量機的結果互相比較。實驗結果顯示，貝氏演算法的正確率為70.67%、支援向量機得到66.01%、深度學習則為89.94%。從而證明深度學習在情感分析上的預測效果最為出色。

The rise of social networking, many consumers are willing to discuss in the community media to share, express their views on the product. Enterprises can analyze the consumers' preferences and advantages and disadvantages of the various products on the market through a large number of online reviews, but in the past the literature is less applied to the Deep Learning in the Sentiment Analysis of Chinese comments.
    The contribution of this thesis is to construct a sentiment dictionary which belongs to the field of Smart Bracelet. And applying Deep Learning and Recursive Neural Network Long Short Memory technology in the Smart Bracelet word of mouth Sentiment Analysis. And compared with the results of Naïve Bayes algorithm and Support Vector Machine. The experimental results show that the correct rate of Naïve Bayes algorithm is 70.67%, the Support Vector Machine is 66.01%, and Deep Learning is 89.94%. So as to prove Deep Learning in the Sentiment Analysis of the most effective prediction.

第一章　緒論	1
1.1	研究背景	1
1.2	研究動機	2
1.3	研究目的	3
1.4	論文架構	4
第二章　文獻探討	6
2.1	文字探勘	6
2.2	情感分析	7
2.3	深度學習	8
2.3.1	遞迴神經網路RNN	9
2.3.2	LSTM網路	11
2.4	智慧型穿戴裝置	12
2.5	本章小結	14
第三章　研究方法與系統架構	16
3.1	研究方法	16
3.2	研究架構	19
3.3	商品類型與語料蒐集	20
3.4	斷詞	22
3.5	評論留言資料收集	23
3.6	資料前處理	26
3.7	新詞萃取	28
3.8	貝氏演算法(Naïve Bayes)	30
3.9	特徵值產生	30
3.10	支援向量機(Support Vector Machine)	32
3.11	產生詞向量(Word Embedding)	34
3.12	深度學習(Deep Learning)	38
3.12.1	活化函數(Activation Function)	41
3.12.2	損失函數(Loss Function)	43
3.12.3	優化器(Optimizers)	44
第四章　資料分析與實驗結果	46
4.1	實驗資料分配與評估方式	46
4.1.1	實驗資料	46
4.1.2	實驗評估方式	47
4.2	基於貝氏演算法(Naïve Bayes)的情感分析	48
4.3	基於支援向量機(Support Vector Machine)的情感分析	48
4.4	基於深度學習遞迴神經網路長短期記憶(Long Short Term Memory)的情感分析	49
4.5	實驗結果	62
第五章　結論與建議	63
5.1	結論	63
5.2	研究貢獻	64
5.3	管理意涵	64
5.4	未來研究方向	65
參考文獻	67
附錄	72

表目錄
=======================================
表 1 結巴斷詞範例	23
表 2 爬蟲程式擷取評論筆數	28
表 3 NTUSD情感詞範例	28
表 4 HowNet情感詞範例	29
表 5 iTSBSD情感詞範例	29
表 6 特徵詞列表	32
表 7 評論語句轉換詞向量陣列範例	37
表 8 詞向量高相似詞彙範例	38
表 9 貝氏演算法預測模型確準率	48
表 10 SVM交叉驗證結果分析	49
表 11 深度學習實驗使用電腦規格	49
表 12 CPU與GPU執行速度比較	50
表 13 遞迴式類神經網路LSTM預測模型使用參數	52
表 14 應用深度學習於智慧型手環情感準確率比較圖	52
表 15 活化函數與各損失函數、優化器最佳搭配比較圖	59
表 16 損失函數與各活化函數、優化器最佳搭配比較圖	60
表 17 優化器與各活化函數、損失函數最佳搭配比較圖	61

圖目錄
=======================================
圖 1 應用深度學習於智慧型手環口碑情感分析研究之研究架構流程	5
圖 2 深度學習神經網路基本架構圖	9
圖 3 RNN架構機制	10
圖 4 LSTM架構圖	11
圖 5 穿戴式裝置示意圖	12
圖 6 穿戴式設備主要功能的演進過程	14
圖 7 系統開發的研究生命週期循環圖	17
圖 8 系統發展研究方法論流程圖	18
圖 9 應用深度學習於智慧型手環口碑情感分析研究系統架構	19
圖 10 小米手環2 產品簡介	21
圖 11 Garmin vivo Smart HR簡介	21
圖 12 華為 TalkBand B3 產品簡介	22
圖 13 淘寶網商品搜尋功能	24
圖 14 天貓商城評價方式	24
圖 15 淘寶集市評價方式	25
圖 16 淘寶集市評論留言	25
圖 17 擷取網頁評論資料流程圖	26
圖 18 搜尋商品總覽網頁資料	27
圖 19 Parser程式解析之後的結構化評論留言資訊	27
圖 20 新詞萃取流程	29
圖 21 貝氏演算法分類與預測資料方法流程	30
圖 22 標記正向語料庫特徵極性	31
圖 23 標記負向語料庫特徵極性	31
圖 24 LibSVM訓練資料方法流程	33
圖 25 10-Fold Cross-Validation測試流程	34
圖 26 Word2vec語料轉換詞向量流程圖	35
圖 27 CBOW演算法模型	36
圖 28 Skip-Gram 演算法模型	36
圖 29 程式執行評論語句轉換成數字陣列範例圖	37
圖 30 程式執行數字陣列轉換成詞向量陣列範例圖	38
圖 31 深度神經網路示意圖	39
圖 32 Keras訓練模型測試程式執行畫面	40
圖 33 應用深度學習訓練之方法流程	41
圖 34 Sigmoid活化函數圖	42
圖 35 TanH活化函數圖	42
圖 36 ReLU活化函數圖	43
圖 37 實驗用語料集資料分配	47
圖 38 以CPU執行MNIST程式處理時間	51
圖 39 以GPU執行MNIST程式處理時間	51
圖 40 深度學習各函數組合的準確率與損失值對比圖	53
圖 41 準確率第一名之訓練曲線圖	54
圖 42 準確率第二名之訓練曲線圖	54
圖 43 準確率第三名之訓練曲線圖	55
圖 44 準確率第四名之訓練曲線圖	55
圖 45 準確率第五名之訓練曲線圖	56
圖 46 未執行Dropout前之深度學習神經網路示意概念圖	58
圖 47 執行Dropout後之深度學習神經網路示意概念圖	58
圖 48 活化函數最佳搭配的準確率與損失值對比圖	59
圖 49 損失函數最佳搭配的準確率與損失值對比圖	60
圖 50 優化器最佳搭配的準確率與損失值對比圖	61
圖 51 應用深度學習於智慧型手環口碑情感分析準確率比較	62

[1]Chang, C.-C., & Lin, C.-J. (2011). LIBSVM: a library for support vector machines. ACM Transactions on Intelligent Systems and Technology (TIST), 2(3), 27. 
[2]Chatterjee, P. (2001). Online reviews: do consumers use them? 
[3]Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078. 
[4]Deng, L., & Yu, D. (2014). Deep learning: methods and applications. Foundations and Trends® in Signal Processing, 7(3–4), 197-387. 
[5]Ding, X., Liu, B., & Yu, P. S. (2008). A holistic lexicon-based approach to opinion mining. Paper presented at the Proceedings of the 2008 international conference on web search and data mining.
[6]Dong, Z., Dong, Q., & Hao, C. (2006). HowNet and the Computation of Meaning. 
[7]Duchi, J., Hazan, E., & Singer, Y. (2011). Adaptive subgradient methods for online learning and stochastic optimization. Journal of Machine Learning Research, 12(Jul), 2121-2159. 
[8]East, R., Hammond, K., & Lomax, W. (2008). Measuring the impact of positive and negative word of mouth on brand purchase probability. International journal of research in marketing, 25(3), 215-224. 
[9]Fan, W., & Gordon, M. D. (2014). The power of social media analytics. Communications of the ACM, 57(6), 74-81. 
[10]Fazayeli, F. (2014). Adaptive Subgradient Methods for Online Learning and Stochastic Optimization. 
[11]Feldman, R., & Sanger, J. (2007). The text mining handbook: advanced approaches in analyzing unstructured data: Cambridge university press.
[12]Graves, A., Liwicki, M., Fernández, S., Bertolami, R., Bunke, H., & Schmidhuber, J. (2009). A novel connectionist system for unconstrained handwriting recognition. IEEE transactions on pattern analysis and machine intelligence, 31(5), 855-868. 
[13]Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. science, 313(5786), 504-507. 
[14]Hinton, G. E., Srivastava, N., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. R. (2012). Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580. 
[15]Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780. 
[16]Ingvaldsen, J. E., & Gulla, J. A. (2012). Industrial application of semantic process mining. Enterprise Information Systems, 6(2), 139-163. 
[17]Jozefowicz, R., Zaremba, W., & Sutskever, I. (2015). An empirical exploration of recurrent network architectures. Paper presented at the Proceedings of the 32nd International Conference on Machine Learning (ICML-15).
[18]Kaur, A., & Duhan, N. (2015). A survey on sentiment analysis and opinion mining. International Journal of Innovations & Advancement in Computer Science, 4, 107-116. 
[19]Kingma, D., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980. 
[20]Ku, L. W., & Chen, H. H. (2007). Mining opinions from the Web: Beyond relevance retrieval. Journal of the American Society for Information Science and Technology, 58(12), 1838-1850. 
[21]LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. 
[22]Mahajan, V., Muller, E., & Kerin, R. A. (1984). Introduction strategy for new products with positive and negative word-of-mouth. Management Science, 30(12), 1389-1404. 
[23]McGonagle, J. J., & Vella, C. M. (2002). A case for competitive intelligence. Information Management, 36(4), 35. 
[24]Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013). Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. 
[25]Money, R. B. (2004). Word-of-mouth promotion and switching behavior in Japanese and American business-to-business service clients. Journal of Business Research, 57(3), 297-305. 
[26]Mullen, T., & Collier, N. (2004). Sentiment Analysis using Support Vector Machines with Diverse Information Sources. Paper presented at the EMNLP.
[27]Nunamaker Jr, J. F., Chen, M., & Purdin, T. D. (1990). Systems development in information systems research. Journal of management information systems, 7(3), 89-106. 
[28]Olah, C. (2015). Understanding lstm networks. GITHUB blog, posted on August, 27, 2015. 
[29]Pang, B., Lee, L., & Vaithyanathan, S. (2002). Thumbs up?: sentiment classification using machine learning techniques. Paper presented at the Proceedings of the ACL-02 conference on Empirical methods in natural language processing-Volume 10.
[30]Popescu, A.-M., Nguyen, B., & Etzioni, O. (2005). OPINE: Extracting product features and opinions from reviews. Paper presented at the Proceedings of HLT/EMNLP on interactive demonstrations.
[31]Raschka, S. (2015). Python machine learning: Packt Publishing Ltd.
[32]Robertson, S. E., & Jones, K. S. (1976). Relevance weighting of search terms. Journal of the American Society for Information science, 27(3), 129-146. 
[33]Singh, J. (1990). Voice, exit, and negative word-of-mouth behaviors: An investigation across three service categories. Journal of the academy of Marketing Science, 18(1), 1-15. 
[34]Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. The Journal of Machine Learning Research, 15(1), 1929-1958. 
[35]Sullivan, D. (2001). Document warehousing and text mining: techniques for improving business operations, marketing, and sales: John Wiley & Sons, Inc.
[36]Tieleman, T., & Hinton, G. (2012). Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude. COURSERA: Neural networks for machine learning, 4(2). 
[37]Tomar, D., Singhal, S., & Agarwal, S. (2014). Weighted least square twin support vector machine for imbalanced dataset. International Journal of Database Theory and Application, 7(2), 25-36. 
[38]Yu, H., & Hatzivassiloglou, V. (2003). Towards answering opinion questions: Separating facts from opinions and identifying the polarity of opinion sentences. Paper presented at the Proceedings of the 2003 conference on Empirical methods in natural language processing.
[39]Zaccone, G. (2016). Getting Started with TensorFlow: Packt Publishing Ltd.
[40]Zeiler, M. D. (2012). ADADELTA: an adaptive learning rate method. arXiv preprint arXiv:1212.5701. 
[41]Zhang, Z., Ye, Q., Zhang, Z., & Li, Y. (2011). Sentiment classification of Internet restaurant reviews written in Cantonese. Expert Systems with Applications, 38(6), 7674-7682. 
[42]財團法人資訊工業策進會. (2014). 穿戴式裝置引領產業新趨勢. 
[43]陳根, & 電腦科學. (2014). 奇" 機" 上身. 第四波工業革命來襲: 穿戴式裝置時代: 上奇時代.
[44]陳智揚. (2015). 穿戴式科技: 科技發展.