近年來，社群網路的蓬勃發展，使得人們除了可以面對面溝通外，還可以利用社群媒體作為電子口碑(E-WOM)傳遞之平台，多元地接收商品訊息。消費者也利用此一管道，對商品或是服務進行了解。其中透過了解負面的電子口碑，可以得知消費者的需求，並提供公司進行產品和服務上的改善以及設計個人化產品的方向。而相關研究亦證實負面的網路評論在於影響電子口碑的重要因素之一，因此如何偵測出負面評論就成為一項重要的議題。然而，當負(正)面評論多於負(正)面的評論時，就會造成分類器對多數類別範例有較高分類正確率，卻對少數類別資料有較低的偵測率，而少數類別往往是較為重要的，這就是所謂的”類別不平衡問題”(class imbalance problems)。資料探勘領域的學者逐漸重視這個問題，並提出各種處理類別不平衡的方法，如重新抽樣、特徵選取等等，使資料類別分佈可以重新取得平衡，以改善分類器對少數類別的分類效能。雖然重新抽樣的方式可以快速處理類別不平衡樣本的問題，但卻缺乏有系統的依據進行分類等等，使改善分類器(如：支持向量機、決策樹)擁有較好的分類效能，因此本研究提出「植基於倒傳遞類神經網路方法」，以倒傳遞類神經網路為基礎作為資料重新抽樣的依據，並與其他傳統處理不平衡分類的方法做比較，以證實所提方法之有效性。

In recent years, the rapid growth of social networks makes people not only can communicate face to face, but also can use social media as platforms for delivering electronic word of mouth (e-WOM), and receiving various product information. Customers also use these channels to get better understandings of services or products. Through obtaining negative e-WOMs, we can understand the voice of customers and thus provide them to companies for improving services and products. Negative e-WOMs also can assist us to design customized products and predict the directions of improvements. Related works also confirmed that negative e-WOM is one of important factors of influencing e-WOM distribution. Consequently, how to detect negative e-WOMs has become one of crucial issues. However, when the amount of positive e-WOMs is larger than the amount of negative e-WOMs, the classifiers induced from the collected comments will have high accuracy for the majority examples, but unacceptable error for the minority examples which are usually the important class. This is so-called “class imbalance problems”. Recently, lots of related scholars have paid attentions to this issue and presented many solutions including re-sampling, feature selection, and so on to re-balance the imbalanced class distribution to improve the classification performance. Since re-sampling methods can quickly balance the imbalance situation, but it lacks a systematic treatment of data to improve the performance of classifiers such as Support Vector Machines (SVM) and Decision Trees (DT). This study proposed the “Back-propagation Neural Network based Approach” to be the basis for re-sampling. In order to testify the effectiveness of the proposed method, the results will be compared with other traditional methods.

目錄
中文摘要 I
Abstract II
致謝 IV
第1章 導論 1
1.1研究背景 1
1.2 研究動機 2
1.3 研究目的 4
第2章 文獻探討 5
2.1 網路口碑 5
2.2 不平衡語意分類 7
2.3 支持向量機 14
2.4 決策樹 16
2.5 類神經網路 18
第3章 研究方法 21
第4章 實驗結果與分析 30
4.1 資料來源與收集 30
4.2 資料前處理 33
4.3 評估指標 35
4.4 實驗結果 37
第5章 結論 66
參考文獻 68
附錄A-資料集三之SVM分類器各實驗結果 75
附錄B -資料集三之DT分類器各實驗結果 81



表目錄
表2-1 處理不平衡分類的相關研究 11
表3-1 詞彙文件矩陣範例 23
表4-1 實驗資料集 32
表4-2 資料集前處理與5 Fold結果 34
表4-3 混淆矩陣 35
表4-4樣本排序示意表 38
表4-5 抽取較小的Yj值之示意圖 39
表4-6 抽取較大的Yj值之示意圖 40
表4-7 資料集一之實驗方法結果比較表 41
表4-8 SVM分類器之調整錯誤分類成本的結果 43
表4-9 實驗結果比較表 45
表4-10 調整錯誤分類成本之DT分類器分類結果 46
表4-11 資料集二之實驗方法比較結果 48
表4-12 調整錯誤分類成本的結果 49
表4-13 方法比較之DT分類結果 51
表4-14 資料集二之調整錯誤分類成本法(DT) 52
表4-15 資料集三實驗總表1(SVM) 54
表4-16 資料集三實驗總表2(SVM) 55
表4-17 資料集三實驗總表3 (SVM) 56
表4-18 資料集三實驗總表4 (SVM) 56
表4-19 資料集三實驗總表5(DT) 57
表4-20 資料集三實驗總表6 (DT) 58
表4-21 資料集三實驗總表7 (DT) 58
表4-22 資料集三實驗總表8 (DT) 59
表4-23 實驗結果總表1(SVM) 60
表4-24 實驗結果總表2(SVM) 61
表4-25 實驗結果總表3(SVM) 62
表4-26 實驗結果總表4(DT) 63
表4-27 實驗結果總表5(DT) 64
表4-28 實驗結果總表6(DT) 65

圖目錄
圖2-1 SVM處理不平衡語意分類示意圖 15
圖3-1 本研究之實驗流程圖 21
圖3-2 隨機減少多數法(Under-sampling)範例 26
圖3-3隨機增加少數法(Over-sampling)範例 27
圖4-1旅遊網站評論的頁面 31
圖4-2 評論中星級的依據 31
圖4-3 資料集一之實驗方法結果比較(SVM) 42
圖4-4 調整錯分成本法之比較結果(SVM) 44
圖4-5 資料集一之實驗方法結果比較(DT) 45
圖4-6 資料集一之調整錯誤分類成本分類結果(DT) 47
圖4-7 資料集二之實驗方法結果比較(SVM) 48
圖4-8 資料集二之調整錯分成本法(SVM) 50
圖4-9 資料集二之實驗方法結果比較(DT) 51
圖4-10 資料集二之調整錯分成本法比較(DT) 53

中文文獻
[1]楊台寧、謝秉訓，民101，『電子口碑、品牌形象、品牌忠誠度與消費者購買意願關係之研究』，第15屆科技整合管理研討會，第1~15頁。
[2]李唯菁，民101，探索電子口碑對購買意圖影響之研究，國立台北科技大學資訊與運籌管理研究所。
[3]蕭涵中，民102，電子口碑訊息來源可信度與產品涉入程度對旅遊產品購買意圖影響之研究，中國文化大學商學院觀光產業學系。
英文文獻
[4]Bao, T., and Chang, T. L. S., 2014. “Finding disseminators via electronic word of mouth message for effective marketing communications,” Decision Support Systems, 67, pp.21-29.
[5]Cheung, C.M.L., Lee, M.K.O., and Thadani, D.R., 2008. “The impact of electronic word-of-mouth: The adoption of online opinions in online customer communities,” Internet Research18(3), pp.229-247.
[6]Chandra, B., Paul Varghese, P., 2009. “Moving towards efficient decision tree construction,” Information Sciences (179), pp.1059–1069.
[7]Chen Y. L., Hung L. T., 2009, “Using decision trees to summarize associative classification rules,” Expert Systems with Applications (36), pp. 2338–2351.
[8]Chen L. S., Liu C.H., Chiu H. J., 2011, “A neural network based approach for sentiment classification in the blogosphere,” Journal of Informetrics (5), pp. 313–322.
[9]Christy M.K. Cheung, Matthew K.O. Lee, 2012. “What drives consumers to spread electronic word of mouth in online consumer-opinion platforms,” Decision Support Systems (53), pp.218-225.
[10]Cateni, S., Colla, V., and Vannucci, M. 2014. “A method for resampling imbalanced datasets in binary classification tasks for real-world problems,” Neurocomputing, 135, pp.32-41.
[11]Chang, H. H., & Wu, L. H., 2014. An examination of negative e-WOM adoption: Brand commitment as a moderator. Decision Support Systems, 59, pp.206-218.
[12]Eric W.K. See-To, Kevin K.W. Ho, 2014. “Value co-creation and purchase intention in social network sites: The role of electronic Word-of-Mouth and trust-A theoretical analysis,” Computers in Human Behavior (31), pp.182-189.
[13]Fu, J., & Lee, S., 2013. “Certainty-based active learning for sampling imbalanced datasets,” Neurocomputing, 119, pp.350-358.
[14]Floyd, K., Freling, R., Alhoqail, S., Cho, H. Y., and Freling, T., 2014. “How Online Product Reviews Affect Retail Sales: A Meta-analysis,” Journal of Retailing(90), pp.217-232.
[15]Farid, D. M., Zhang, L., Rahman, C. M., Hossain, M. A., and Strachan, R., 2014. “Hybrid decision tree and naïve Bayes classifiers for multi-class classification tasks,” Expert Systems with Applications, 41(4), pp.1937-1946.
[16]Farquad, M. A. H., Ravi, V., and Raju, S. B., 2014. “Churn prediction using comprehensible support vector machine: An analytical CRM application,” Applied Soft Computing, 19, pp.31-40.
[17]García, V., Sánchez, J. S., and Mollineda, R. A. 2011. “On the effectiveness of preprocessing methods when dealing with different levels of class imbalance,” Knowledge-Based Systems, vol. 25, no. 1, pp. 13-21.
[18]Gupta, P., and Harris, J., 2010. “How e-WOM recommendations influence product consideration and quality of choice: a motivation to process information perspective,” Journal of Business Research, 63(9), 1041-1049.
[19]Hao, M., Wang, Y., & Bryant, S. H., 2014. “An efficient algorithm coupled with synthetic minority over-sampling technique to classify imbalanced PubChem BioAssay data,” Analytica chimica acta, 806, pp.117-127.
[20]Hogenboom, A., Heerschop, B., Frasincar, F., Kaymak, U., & de Jong, F., 2014. “Multi-lingual support for lexicon-based sentiment analysis guided by semantics,” Decision Support Systems, 62, pp.43-53.
[21]Ilyoo B. Hong, Hoon S. ChaCollege, 2013. “The mediating role of consumer trust in an online merchant inpredicting purchase intention,” International Journal of Information Management (33), pp.927– 939.
[22]Kuo R.J., Shieh M.C., Zhang J.W., Chen K.Y., 2013, “The application of an artificial immune system-based back-propagation neural network with feature selection to an RFID positioning system,” Robotics and Computer-Integrated Manufacturing(29), pp.431–438
[23]Krawczyk, B., Wozniak, M., and Schaefer, G., 2014. “Cost-sensitive decision tree ensembles for effective imbalanced classification,” Applied Soft Computing, 14, pp.554-562.
[24]Kang, S., and Cho, S., 2014. “Approximating support vector machine with artificial neural network for fast prediction,” Expert Systems with Applications, 41(10), pp.4989-4995.
[25]Li, S., Zhou, G., Wang, Z., Lee, S. Y. M., and R. Wang, 2011. “Imbalanced sentiment classification,” Proceedings of the 20th ACM International Conference on Information and Knowledge Management, pp.2469-2472.
[26]Liu, Y., Yu X., Huang, J. X., An A. 2011. “Combining integrated sampling with SVM ensembles for learning from imbalanced datasets,” Information Processing and Management(47), pp.617-631.
[27]Li, Z., Yang, X., Gu, W., Zhang, H., 2013, “Kernel-improved Support Vector Machine for semanteme data,” Applied Mathematics and Computation (219), pp.8876–8880.
[28]Lopez, F. M., Puertas, S. M., and Arriaza, J. T. 2014. Training of support vector machine with the use of multivariate normalization. Applied Soft Computing, 24, 1105-1111.
[29]Liao, J. J., Shih, C. H., Chen, T. F., and Hsu, M. F., 2014. “An ensemble-based model for two-class imbalanced financial problem,” Economic Modelling, 37, pp.175-183.
[30]Lam, H. K., Ekong, U., Liu, H., Xiao, B., Araujo, H., Ling, S. H., and Chan, K. Y., 2014. “A Study of Neural-Network-Based Classifiers for Material Classification,” Neurocomputing.
[31]Moraes R., Valiati J. F., Wilson P., Neto, G. 2013. “Docnment-level sentiment classification: An empirical comparison between SVM and ANN,” Expert Systems with Applications(40), pp.621–633.
[32]Meng, J., Lin, H., and Yu, Y. (2011), “A two-stage feature selection method for text categorization,” Computers and Mathematics with Applications, vol. 62, no. 7, pp. 2793-2800.
[33]Mohd Nawi, N., Atomia, W. H., and Rehman, M. Z. (2013). “The effect of data pre-processing on optimized training of artificial neural networks,” Procedia Technology (11), pp.32 –39.
[34]Orriols-Puig, A., and Bernadó-Mansilla, E., 2009. “Evolutionary rule-based systems for imbalanced data sets,” Soft Computing, 13(3), pp.213-225.
[35]Polat, K., and Güneş, S., 2009. “A novel hybrid intelligent method based on C4. 5 decision tree classifier and one-against-all approach for multi-class classification problems,” Expert Systems with Applications, 36(2), pp.1587-1592.
[36]Pai, M. Y., Chu, H. C., Wang, S. C., & Chen, Y. M., 2013. “Electronic word of mouth analysis for service experience,” Expert Systems with Applications,40(6), pp.1993-2006.
[37]Qian, Y., Liang, Y., Li, M., Feng, G., and Shi, X., 2014. “A resampling ensemble algorithm for classification of imbalance problems,” Neurocomputing, 143, pp.57-67.
[38]Rodrigo M., Valiati J. F., Wilson P. G. N., 2013, “Document-level sentiment classification: An empirical comparison between SVM and ANN,” Expert Systems with Applications (40), pp. 621–633.
[39]Rutkowski, L., Jaworski, M., Pietruczuk, L., and Duda, P., 2014. “The CART decision tree for mining data streams,” Information Sciences, 266, pp.1-15.
[40]Rumelhart, D.E., McClelland, J.L., and the PDP Research Group,1986.PARALLEL DISTRIBUTED PROCESSING, Vol. 1, MIT Press, Cambridge,MA.
[41]Ramedani, Z., Omid, M., Keyhani, A., Shamshirband, S., and Khoshnevisan, B., 2014. “Potential of radial basis function based support vector regression for global solar radiation prediction,” Renewable and Sustainable Energy Reviews,39, pp.1005-1011.
[42]Sun, A., Lim E.P., Liu Y. 2009. "On strategies for imbalanced text classification using SVM: A comparative study," Decision Support Systems 48 pp.191-201.
[43]Tong, L. I., Chang, Y. C., and Lin, S. H., 2011, “Determining the optimal re-sampling strategy for a classification model with imbalanced data using design of experiments and response surface methodologies,” Expert Systems with Applications, vol. 38, no. 4, pp. 4222-4227.
[44]Verhagen, T., Nauta A., Feldberg F. 2013. “Negative online word-of-mouth:Behavioral indicator or emotional release?,” Computers in Human Behavior 29, pp.1430–1440.
[45]Vapnik, V. N. 1995. “The Nature of Statistical Learning Theory,” Springer-Verla.
[46]Verhagen, T., Nauta, A., and Feldberg, F., 2013. “Negative online word-of-mouth: Behavioral indicator or emotional release?,” Computers in Human Behavior, 29(4), pp.1430-1440.
[47]Wang, R., Kwong, S., Chen, D., “Inconsistency-based active learning for support vector machines,” 2012, Pattern Recognition (45), pp.3751–3767.
[48]Wu, Q., Ye, Y., Zhang, H., Ng, M. K., and Ho, S. S., 2014. “ForesTexter: An efficient random forest algorithm for imbalanced text categorization,” Knowledge-Based Systems, 67, pp.105-116.
[49]Yu H., Ni j., Zhao J., 2013. ” ACOSampling: An ant colony optimization-based undersampling method for classifying imbalanced DNA microarray data,” Neurocomputing (101), pp.309–318.
[50]Yang J., Liu Y., Zhu X., Liu Z., Zhang X. 2012. “A new feature selection based on comprehensive measurement both in inter-category and intra-category for text categorization,” Information Processing and Management (48), pp.741–754.
[51]Yoo, C. W., Sanders, G. L., & Moon, J., 2013. “Exploring the effect of e-WOM participation on e-Loyalty in e-commerce,” Decision Support Systems, 55(3), pp.669-678.
[52]Yin, L., Ge, Y., Xiao, K., Wang, X., and Quan, X., 2013. “Feature selection for high-dimensional imbalanced data,” Neurocomputing, 105, pp.3-11.
[53]Zhang, H., and Li, M., 2014. “RWO-Sampling: A random walk over-sampling approach to imbalanced data classification. Information Fusion,” 20, pp.99-116.
[54]Zhao, Z., Zhong, P., and Zhao, Y. (2011). Learning SVM with weighted maximum margin criterion for classification of imbalanced data. Mathematical and Computer Modelling, 54(3), pp.1093-1099.
[55]Zhang, H., and Li, M., 2014. “RWO-Sampling: A random walk over-sampling approach to imbalanced data classification,” Information Fusion, 20, pp.99-116