臺灣博碩士論文加值系統

在資料探勘(Data Mining)與機器學習(Machine learning)的領域中，分類(Classification)技術受到相當廣泛的研究，並應用於各領域的實務中。一般來說，當訓練資料集中的目標類別分佈越均勻時，分類器(Classifier)通常會有比較好的分類效能。然而，在許多實務的應用中，卻經常牽涉到目標類別分佈不平衡的問題(Imbalanced Class Distribution Problem)。目標類別分佈不平衡所造成的問題是，當大多數的訓練資料是屬於某一類別(我們稱之為多數類別(Majority Class))，而只有極少數的訓練資料是屬於另一個類別(我們稱之為少數類別(Minority Class))時，其所訓練出來的分類器通常會傾向將所有的測試資料預測為多數類別，完全忽略少數類別，而少數類別的預測效能又是決策者最重視的部分。因此對於不平衡資料探勘進行研究就顯得相當重要。本論文將提出一個過濾多數類別資料的方法流程，利用類神經網路模型把預測為高機率值的多數類別資料過濾掉，進而提高少數類別資料所佔的比例，以降低資料不平衡的程度，藉此達到多數類別資料與少數類別資料的平衡。之後分為兩個方法流程：流程一為利用集群分析把資料分多群並且每群各自建立分類模型；流程二則不分群而直接建立分類模型，並且對分類模型做效能最佳化。實驗的結果顯示我們所提出的方法確實能提昇傳統不平衡資料之分類預測方法的分類效能。

In real data, the data distribution is imbalanced when the quantity of some classes is usually much less than other classes and it is called ‘Imbalance data’. The data in minority classes are quite important in research. In the classification technique of data mining, the training data quality is a critical factor which can influence the accuracy of the technique. However, traditional data mining classification technique is not effective on imbalance data. Hence, it is a quite significant goal to improve the performance of result on mining imbalance data. This paper will present an approach that filter majority class by neural networks classification model. The neural networks classification model will be used to filter majority class which was high probability forecast for majority class. After that, the ration of data in minority of classes will be increased, and the extent of imbalanced class distribution will be decreased. Then we made two different methods: Method 1 will use the cluster analysis to segment the data into multiple groups and the data in each group will be used to build the classification model; Method 2 will use the data to directly build the classification model instead of segment the data into multiple groups. At the same time, we will optimize the performance of the classification model. Finally, the different sampling technique will be used to select the data from well-handled dataset and classification model will be built up. The experiments show that our approach can increase the performance of traditional classification technique in imbalance dataset.

中文摘要 I
英文摘要 II
致謝 III
目錄 IV
表目錄 VI
圖目錄 VIII
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 1
1.3 研究目的 2
第二章 相關研究 4
2.1 不平衡資料集 4
2.2 不平衡資料集所引起的問題 4
2.3 解決不平衡資料集探勘的方法 5
2.4 資料探勘技術 6
2.4.1 分類模型: 7
2.4.2 抽樣技術 9
第三章 研究方法 22
3.1 過濾多數類別資料後分群 22
3.2 過濾多數類別資料後不分群 23
3.3屬性重要度分析 24
3.3.1 數值屬性分析方法 24
3.3.2 類別屬性分析方法 25
3.4 門檻值的訂定 27
3.5 模型效能最佳化 27
3.5 抽樣技術 30
第四章 本研究實驗 31
4.1 中小企業資料集 32
4.1.1 效能評估 34
4.2 信用風險資料集 38
4.2.1 效能評估 39
4.3 房貸資料集 43
4.3.1 效能評估 45
第五章 結論與未來工作 50
參考文獻 51
附錄 54
附錄A: SME中小企業資料集 54
附錄B: 房貸資料集 58
附錄C: 信用風險資料集 62

[1]譚琳，“非平衡數據挖掘簡介”，計算機科學與技術研討會，南京，2008。
[2]張琦、吳斌、王柏，“非平衡數據訓練方法概述”，計算機科學研討會，北京，2005。
[3]K. E. Fish, J. H. Barnes, M. W. Aiken, “Artificial Neural Networks: a New Methodology for Industrial Market Segmentation.” Industrial Marketing Management, Vol. 24, Issue 5, pp. 431-438, 1995.
[4]Peter E. Hart. “The Condensed Nearest Neighbour Rule.” IEEE Transactions on Information Theory, Vol. 14, Issue 3, pp. 515-516, 1968.
[5]Ivan Tomek, “Two Modifications of CNN,” IEEE Transactions on Systems Man and Communications, Vol. 6, Issue 11, pp. 769-772, 1976.
[6]Jianping Zhang, Inderjeet Mani, “KNN Approach to Unbalanced Data Distributions: A Case Study involving Information Extraction,” Proceedings of International Conference on Machine Learning Workshop on Learning from Imbalanced Data Sets (II), 2003.
[7]Show-Jane Yen and Yue-Shi Lee, “Cluster-based Under-Sampling Approaches for Imbalanced Data Distributions,” Expert Systems with Applications, Vol. 36, Issue 3, pp. 5718-5727, 2009.
[8]Yue-Shi Lee, Show-Jane Yen and Shin-Mau Chen, "Combine Cluster Analysis to Classification Prediction in Imbalanced Data Distribution," Proceedings of International Symposiums on Major Academic Disciplines, pp. 111-122, 2009.
[9] Nitesh V. Chawla, Kevin W. Bowyer, Lawrence O. Hall, W. Philip Kegelmeyer, “SMOTE: Synthetic Minority Over-Sampling Technique,” Journal of Artificial Intelligence Research, Vol.16, pp. 321-357, 2002.
[10]Tian-Shyug Lee, Chih-Chou Chiu, Yu-Chao Chou, Chi-Jie Lu, “Mining the Customer Credit Using Classification and Regression Tree and Multivariate Adaptive Regression Splines,” Computational Statistics and Data Analysis, Vol. 50, Issue 4, pp. 1113-1130, 2006.
[11]Xiu Li, Weiyun Ying, Jianyong Tuo, Bing Li, Wenhuang Liu, “Applications of classification trees to consumer credit coring methods in commercial banks,” IEEE International Conference on Systems, Man and Cybernetics, Vol. 5, Issue 10, pp.4112-4117, 2004.
[12]Vellido A., Lisboa P.J.G., Vaughan J., “Neural Networks in Business: A Survey of Applications(1992-1998),” Expert System With Applications, Vol. 17, Issue 1, pp.51-70, 1999.
[13]Chih-Chou Chiu, Yuehjen E. Shao, Tian-Shyug Lee, Ker-Ming Lee, ”Identification of Process Disturbance Using SPC/EPC and Neural Networks,” Journal of Intelligent Manufacturing, Vol.14, Issue 13, pp. 379-388, 2003.
[14]Tian-Shyug Lee, Chih-Chou Chiu, Chi-Jie Lu, I-Fei Chen, “Credit Scoring Using the Hybrid Neural Discriminate Technique,” Expert Systems with Applications, Vol. 23, Issue 3, pp. 245-254, 2002.
[15]Guoqiang Zhang, B. Eddy Patuwo, Michael Y. Hu, “Forecasting with artificial neural networks: The state of the art,” International Journal of Forecasting, Vol. 14, Issue 1, pp.35-62, 1998.
[16]Eliana Angelini, Giacomo di Tollo, Andrea Roli, “A Neural Network Approach for Credit Risk Evaluation,” The Quarterly Review of Economics and Finance, Vol. 48, Issue 4, pp.733-755, 2008.
[17]Tian-Shyug Lee, I-Fei Chen, “A Two-stage Hybrid Credit Scouring Model Using Artificial Neural Networks and Multivariate Adaptive Regression Splines,” Expert Systems with Applications, Vol. 28, Issue 4, pp. 743-752, 2005.
[18]Z. S. Marijana, S. Natasa, B. Mirta, “Small Business Credit Scoring: A comparison of Logistic Regression, Neural Network, and Decision Tree Models,” Proceedings of the 26th International Conference on Information Technology Interfaces, Vol. 1, Issue 10, pp. 265-270, 2004.