臺灣博碩士論文加值系統

在虛擬社群的快速發展下，部落客(Bloggers) 藉由網際網路快速傳遞對於任何產品或服務的評論，而一些負面評論往往會降低消費者的購買意願，並帶給企業嚴重的傷害。然而，這些非結構化的文本資訊，卻難以在短時間內理解。在某些情況下，負面評論通常少於正面評論，這些較少的負面評論傳播快速且具有極大殺傷力。在語意分類(Sentiment Classification)的領域中，學者往往未考慮到這種非等量的類別不平衡問題(Class Imbalance Problems)。而類別不平衡問題是指分類器對多數類別有高分類準確率，但對少數類別則會產生過高的分類錯誤率。因此，從大量的網路評論中偵測出少數的負面評論，已成為重要的議題。本研究目的是確定影響不平衡語意分類準確率的關鍵因子，使用田口方法(Taguchi Methods)以找出重要的影響因子。最後，根據所找出的關鍵因子提出「平衡類別特徵(Balanced Category Features, BCF)」方法，並結合「隱含語意索引(Latent Semantic Index, LSI)」，以改善不平衡語意分類效能。此外，本研究透過支持向量機(Support Vector Machines, SVM)、決策樹(Decision Tree, DT)來建構分類模型，並使用真實的網路評論進行實驗，以證實所提出方法的有效性。

The fast development of virtual Social network, the bloggers send their comments to the products or services quickly via the internet. Some negative comments could reduce consumers’ purchase intentions and bring a great damage to enterprises. However, the text information in blogs are often unstructured and hard to comprehend in short time. In some cases, the negative comments are usually fewer than the positive opinions. These fewer negative comments spread very fast and are much harmful. In the Sentiment classification study, the researchers always didn’t consider the class imbalance problem. A classifier induced from an imbalanced data set has high classification accuracy for the majority class, but an unacceptable error rate for the minority class. Therefore, to identify consumers’ negative sentiments effectively from a large number of online comments had become one of serious issues. So, this study aims to identify the key factors of imbalanced sentiment classification by using Taguchi method. Then, according to the discovered key factors, we proposed a new method “Balanced Category Features (BCF)” and combine the “Latent Semantic Index (LSI)” to improve the performance of the imbalanced sentiment classification. Moreover, Support Vector Machines (SVM) and Decision Tree (DT) have been employed to construct classifiers. Finally, one case study from real world blogs will be provided to illustrate the effectiveness of our proposed approach.

目錄
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 3
1.3 研究目的 6
1.4 研究架構 6
1.5 研究步驟 8
第二章 文獻探討 10
2.1 不平衡語意分類問題 10
2.1.1 語意分類 10
2.1.2 不平衡語意分類 11
2.1.3 小結 18
2.2維度縮減方法 19
2.2.1 特徵選取 19
2.2.2 特徵擷取 27
2.3 田口方法 29
2.4 機器學習方法 36
2.4.1 支持向量機 36
2.4.2 決策樹 38
第三章 研究方法 41
3.1 應用田口方法找出關鍵因子 41
3.2 平衡類別特徵 45
3.3 BCF結合LSI二階段方法 51
第四章 實驗結果與分析 54
4.1 田口方法之實驗結果 54
4.1.1 資料來源與工具 54
4.1.2 資料前處理 55
4.1.3 評估指標 57
4.1.4 實驗結果與分析 59
4.2 BCF+LSI之實驗結果 64
4.2.1 資料來源與工具 64
4.2.2 資料前處理 65
4.2.3 實驗結果 67
第五章 結論與未來研究方向 114
5.1 研究結論 114
5.2 未來研究方向 116
參考文獻 118
圖目錄
圖1 1 不平衡語意分類問題 3
圖1 2 研究架構圖 7
圖1 3 研究步驟 9
圖2 1 處理不平衡問題相關統計 15
圖2 2 特徵型態分佈圖 23
圖2 3 Relief方法之計算原理 25
圖2 4 奇異值分解 29
圖2 5 產品/製程參數圖 31
圖2 6 直交表之符號定義 33
圖2 7 因子效果圖範例 35
圖2 8 SVM超平面示意圖 37
圖3 1 田口方法流程 44
圖3 2 BCF流程圖 47
圖3 3 BCF結合LSI之流程 53
圖4 1 因子效果圖 63
圖4 2 Review centre評論網站五星等評價系統 64
圖4 3 傳統不平衡資料處理方法之實驗結果(SVM) 70
圖4 4 傳統不平衡資料處理方法之實驗結果(DT) 70
圖4 5 傳統不平衡資料處理方法之實驗結果-調整Cost(SVM) 72
圖4 6 傳統不平衡資料處理方法之實驗結果-調整Cost(DT) 72
圖4 7 手機評論之實驗結果(SVM) 75
圖4 8 手機評論之實驗結果-調整Cost(SVM) 78
圖4 9 手機評論之實驗結果(DT) 81
圖4 10 手機評論之實驗結果-調整Cost(DT) 84
圖4 11 iPhone評論之實驗結果(SVM) 87
圖4 12 iPhone評論之實驗結果-調整Cost(SVM) 90
圖4 13 iPhone評論之實驗結果(DT) 93
圖4 14 iPhone評論之實驗結果-調整Cost(DT) 96
圖4 15 相機評論之實驗結果(SVM) 99
圖4 16 相機評論之實驗結果-調整Cost(SVM) 102
圖4 17 相機評論之實驗結果(DT) 105
圖4 18 相機評論之實驗結果-調整Cost(DT) 108

 
表目錄
表2 1 處理不平衡問題相關研究 13
表2 2 特徵選取方法使用於不平衡文本分類之相關研究 20
表2 3 詞彙與類別關係表 22
表2 4 L8(27)直交表 33
表3 1 實驗因子與水準 42
表3 2 Sign指標計算範例 49
表3 3 Sign-IG計算範例 50
表3 4 詞彙文件矩陣範例 51
表4 1 田口方法之資料來源 55
表4 2 田口方法之原始實驗資料 55
表4 3 田口方法之前處理結果 56
表4 4 混淆矩陣 58
表4 5 田口方法之實驗結果 60
表4 6 田口實驗之回應值與SN比(GM指標) 62
表4 7 SN比之ANOVA表 62
表4 8 因子貢獻度與ANOVA表 62
表4 9 實驗資料 65
表4 10 資料集之特徵集合 66
表4 11 平衡後特徵數 67
表4 12 傳統不平衡資料處理方法之實驗結果 69
表4 13 傳統不平衡資料處理方法之實驗結果-調整Cost 71
表4 14 手機評論之實驗結果(SVM) 73
表4 15 手機評論之實驗結果-調整Cost(SVM) 76
表4 16 手機評論之實驗結果(DT) 79
表4 17 手機評論之實驗結果-調整Cost(DT) 82
表4 18 iPhone評論之實驗結果(SVM) 85
表4 19 iPhone評論之實驗結果-調整Cost(SVM) 88
表4 20 iPhone評論之實驗結果(DT) 91
表4 21 iPhone評論之實驗結果-調整Cost(DT) 94
表4 22 相機評論之實驗結果(SVM) 97
表4 23 相機評論之實驗結果-調整Cost(SVM) 100
表4 24 相機評論之實驗結果(DT) 103
表4 25 相機評論之實驗結果-調整Cost(DT) 106
表4 26分類效能最佳與次佳之方法 109
表4 27 傳統方法之訓練時間 110
表4 28 手機評論之訓練時間(SVM) 110
表4 29 手機評論之訓練時間(DT) 111
表4 30 iPhone評論之訓練時間(SVM) 111
表4 31 iPhone評論之訓練時間(DT) 112
表4 32 相機評論之訓練時間(SVM) 112
表4 33 相機評論之訓練時間(DT) 113

中文文獻
[1]曾韋榮 (2006），結合潛在語意檢索及資訊粒化於資料探勘，碩士論文，國立臺北科技大學商業自動化與管理研究所，臺北。
[2]張麗新、王家廞、趙雁南、楊澤紅 (2004)，「基於Rel ief 的組合式特徵選擇」，復旦學報(自然科學版)，第43卷，第5期，第893-898頁。
[3]廖闊、付建勝、楊萬麟 (2010)，「改進的ReliefF 算法用於雷達距離像目標識別」，電子測量與儀器學報，第24卷，第9期，第831-836頁。
英文文獻
[1]Abbasi, A., and Chen, H. (2005), “Applying authorship analysis to extremist-group web forum messages,” IEEE Intelligent Systems, vol. 20, no. 5, pp. 67–75.
[2]Abbasi, A., Chen, H., and Salem, A. (2008), “Sentiment analysis in multiple languages: feature selection for opinion classification in web forums,” ACM Transactions on Information Systems, vol. 26, no. 3, pp. 12:1-12:34.
[3]Arun Kumar, M., and Gopal, M. (2010), “A comparison study on multiple binary-class SVM methods for unilabel text categorization,” Pattern Recognition Letters, vol. 31, no. 11, pp. 1437-1444.
[4]Chang, C. C. and Lin, C. J. (2001), “LIBSVM: A Library for Support Vector Machines,” Software, available at: http://www.csie.ntu.edu.tw/~cjlin/libsvm.
[5]Chaovalit, P., and Zhou, L. (2005), “Movie review mining: a comparison between supervised and unsupervised classification approaches,” In Proceedings of the 38th Hawaii International Conference on System Sciences, pp.1-9.
[6]Chen, E., Lin, Y., Xiong, H., Luo, Q., and Ma, H. (2010), “Exploiting probabilistic topic models to improve text categorization under class imbalance,” Information Processing and Management, vol. 47, no. 2, pp. 202-214.
[7]Chen, L. S., Liu, C. H., and Chiu, H. J. (2011), “A neural network based approach for sentiment classification in the blogosphere,” Journal of Informetrics, vol. 5, no. 2, pp. 313-322.
[8]Chien, W. T., and Tsai, C. S. (2003), “The investigation on the prediction of tool wear and the determination of optimum cutting conditions in machining 17-4PH stainless steel,” Journal of Materials Processing Technology, vol. 140, no. 1-3, pp. 340-345.
[9]Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., and Harshman, R. (1990), “Indexing by latent semantic analysis,” Journal of the American Society for Information Science, vol. 41, no. 6, pp. 391-407.
[10]Das Mohapatra, P.K., Maity, C., Rao, R.S., Pati, B.R., and Mondal, K.C. (2009), “Tannase production by bacillus licheniformis KBR6: optimization of submerged culture conditions by taguchi DOE methodology,” Food Research International, vol. 42, no. 4, pp. 430-435.
[11]Fern&;aacute;ndez, A., del Jesus, M. J., and Herrera, F. (2009), “On the influence of an adaptive inference system in fuzzy rule based classification systems for imbalanced data sets,” Expert Systems with Application, vol. 36, pp. 9805-9812.
[12]Garc&;iacute;a, V., S&;aacute;nchez, 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.
[13]Gunn, S. R. (1998), “Support vector machines for classification and regression,” Technical Report, University of Southampton, UK.
[14]Gomez, J. C., and Moens, M. F. (2012), “PCA document reconstruction for email classification,” Computational Statistics and Data Analysis, vol. 56, no. 3, pp. 741-751.
[15]Huang, Y., McCullagh, P. J., Black, N. D. (2009), “An optimization of ReliefF for classification in large datasets,” Data &; Knowledge Engineering, vol. 68, no. 11, pp. 1348-1356.
[16]Hong, C. W. (2011), “Using the Taguchi method for effective market segmentation,” Expert Systems with Applications, doi:10.1016/j.eswa.2011.11.040.
[17]Kira, K., and Rendell, L. A. (1992), “The feature selection problem traditional methods and a new algorithm,” In Proceedings of 9th National Conference on Artificial Intelligence, pp. 129-134.
[18]Kononenko, I. (1994), “Estimating attributes: analysis and extensions of Relief,” In Proceedings of the European Conference on Machine Learning, pp. 171-182.
[19]Kontostathis, M., and Pottenger, W. M. (2006), “A framework for understanding latent semantic indexing (LSI) performance,” Information Processing and Management, vol. 42, no. 1, pp. 56-73.
[20]Keshtkar, F., and Inkpen, D. (2009), “Using sentiment orientation features for mood classification in blogs,” IEEE International Conference on Natural Language Processing and Knowledge Engineering.
[21]Li, B., Xu, S., and Zhang, J. (2007), “Enhancing clustering blog documents by utilizing author/reader comments,” In Proceedings of the 45th Annual Southeast Regional Conference, pp. 94-99.
[22]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.
[23]Liu, B., Hu, M., and Cheng, J. (2005), “Opinion observer: analyzing and comparing opinions on the web,” In Proceedings of the 14th International Conference on World Wide Web, pp. 342-351.
[24]Liu, Y., Loh, H. T., and Sun, A., (2009), “Imbalanced text classification: a term weighting approach,” Expert Systems with Applications, vol. 37, no. 1, pp. 690-701.
[25]Liu, Y., Yu, X., Huang, J. X., and An, A. (2010), “Combining integrated sampling with SVM ensembles for learning from imbalanced datasets,” Information Processing and Management, vol. 47, no. 4, pp. 617-631.
[26]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.
[27]Mallick, K., and Bhattacharyya, S. (2012), “Uncorrelated local maximum margin criterion: an efficient dimensionality reduction method for text classification,” Procedia Technology, vol. 4, pp. 370-374.
[28]Ogura, H., Amano, H., and Kondo, M. (2010), “Distinctive characteristics of a metric using deviations from Poisson for feature selection,” Expert Systems with Applications, vol. 37, no. 3, pp. 2273–2281.
[29]Ogura, H., Amano, H., and Kondo, M. (2011), “Comparison of metrics for feature selection in imbalanced text classification,” Expert Systems with Applications, vol. 38, no. 5, pp. 4978–4989.
[30]O’Keefe, T., and Koprinska, I. (2009), “Feature selection and weighting methods in sentiment analysis,” In proceedings of the 14th Australasian Document Computing Symposium.
[31]Quinlan, J. R. (1993), “C4.5: programs for machine learning,” Morgan kaufmann, San Mateo, CA.
[32]Sebastiani, F. (2002), “Machine learning in automated text categorization,” ACM Computing Surveys, vol. 34, no. 1, pp. 1–47.
[33]Simeon, M., and Hilderman, R. (2008), “Categorical proportional difference: A feature selection method for text categorization,” In Proceedings of the 17th Australasian Data Mining Conference.
[34]Stamatatos, E. (2008), “Author identification: using text sampling to handle the class imbalance problem,” Information Processing and Management, vol. 44, no. 2, pp. 790-799.
[35]Sun, A., Lim, E. P., and Liu, Y. (2009), “On strategies for imbalanced text classification using SVM: a comparative study,” Decision Support Systems, vol. 48, no. 1, pp. 191-201.
[36]Tan, S., and Zhang, J. (2008), “An empirical study of sentiment analysis for chinese documents,” Expert Systems with Applications, vol. 34, no. 4, pp. 2622-2629.
[37]Tang, H., Tan, S., and Cheng, X. (2009), “A survey on sentiment detection of reviews,” Expert Systems with Applications, vol. 36, no. 7, pp. 10760-10773.
[38]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.
[39]Uğuz, H. (2011), “A two-stage feature selection method for text categorization by using information gain, principal component analysis and genetic algorithm,” Knowledge-Based Systems, vol. 24, no. 7, pp. 1024-1032.
[40]Uysal, A., and Gunal, S. (2012), “A novel probabilistic feature selection method for text classification” Knowledge-Based Systems, doi: 10.1016/j.knosys.2012.06.005.
[41]van Halteren, H. (2004), “Linguistic profiling for author recognition and verification,” Proceedings of the 42nd annual meeting of the association for computational linguistics, pp. 199–206.
[42]Vapnik, V. N. (1995), The Nature of Statistical Learning Theory, Springer-Verlag.
[43]Weiss, G., and Provost, F. (2003), “Learning when training data are costly: the effect of class distribution on tree induction,” Journal of Artificial Intelligence Research, vol. 19, no. 1, pp. 315-354.
[44]Whitelaw, C., Garg, N., and Argamon, S. (2005), “Using appraisal groups for sentiment analysis,” In proceedings of the ACM 14th Conference on Information and Knowledge Management, pp. 625-631.
[45]Wu, C. H., Chuang, Z. J., and Lin, Y. C. (2006), “Emotion recognition from text using semantic labels and separable mixture models,” ACM Transactions on Asian Language Information Processing, vol. 5, no. 2, pp. 165-182.
[46]Ye, Q., Zhang, Z., and Law, R. (2009), “Sentiment classification of online reviews to travel destinations by supervised machine learning approaches,” Expert Systems with Applications, vol. 36, no. 3, pp. 6527-6535.
[47]Yusoff, N., Ramasamy, M., and Yusup, S. (2011), “Taguchi’s parametric design approach for the selection of optimization variables in a refrigerated gas plant,” Chemical Engineering Research and Design, vol. 89, no. 6, pp. 665-675.
[48]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, vol. 48, no. 4, pp. 741-754.
[49]Zhang, J. Z., Chen, J. C., and Kirby, E. D. (2007), “Surface roughness optimization in an end-milling operation using the Taguchi design method,” Journal of Materials Processing Technology, vol. 184, no. 1-3, pp. 233-239.
[50]Zheng, Z., Wu, X., and Srihari, R. (2004), “Feature selection for text categorization on imbalanced data,” ACM SIGKDD Explorations Newsletter, vol. 6, no. 1, pp. 80-89.
[51]Zhang, W., Yoshida, T., and Tang, X. (2011), “A comparative study of TF*IDF, LSI and multi-words for text classification,” Expert Systems with Applications, vol. 38, no. 3, pp. 2758-2765.