近年來，機器學習的發展為分類問題(Classification problems)提供一項有效的分析工具。在監督式學習方法中，支持向量器(Support Vector Machines)是最熱門且具有良好理論基礎的分類器之ㄧ。然而，支持向量器在訓練複雜資料時，很容易受到資料型態的影響，其中複雜的資料型態包含類別不平衡(Class imbalance)、類別重疊(Class overlapping)資料等。當在類別不平衡或高重疊資料中學習時，支持向量器所產生的決策邊界(Decision boundary)會往較大的類別集(Majority class)靠攏趨近；而在預測極少數目標(Rarest objects)時，會忽略少數類別的分類正確率，因此需要發展一個穩定的支持向量器以提升分類能力。此外，由於支持向量器所產生的決策邊界，呈現出複雜的數學模式，缺乏清晰的知識表達(Knowledge representation)能力，因此需要發展一個支持向量器的規則萃取(Rule extraction)演算法來提昇其解釋能力。

本研究目的在於發展出三個模式，“MS-SVM”、“GASVM”、“KCGex-SVM”。MS-SVM模式主要提出修正支持向量器中的差額變數(Slack variable)，以期在複雜的資料中能有好的分類效能；本研究使用人工模擬的資料評估MS-SVM的有效性，在使用不同的績效指標(如Accuracy, Sensitivity和Specificity)評估下，相較於原始的支持向量器，實驗結果可以說明所提方法的優異性。GASVM和KCGex-SVM模式主要用來增加支持向量器的規則萃取能力，本研究使用不同的績效指標(Accuracy, Coverage, Fidelity 和 Comprehensibility)來評估此二個規則萃取模式在UCI資料銀行的資料中的解釋和分類能力，實驗結果指出所提規則萃取模式優於其他著名的規則萃取方法。因此，本研究所提出的規則萃取模式可視為資料挖掘領域中的一種有效分析工具。

本研究採用ㄧ個實際案例來評估所提出模式的有效性，以術後壓瘡(Pressure ulcer)的實際案例，說明MS-SVM和GASVM模式有優異的能力偵測疾病的發生；結果顯示，本研究提出方法於真實案例上確實可行。

Recently, the development of machine-learning techniques has provided an effective analysis tool for classification problems. The support vector machine (SVM) is one of the most popular supervised learning techniques. However, the SVM may not effectively detect the instance of the minority class and obtain lower classification performance in the overlap region when learning from complicated data sets. The complicated data sets with class imbalanced and overlapped distributions are common in most practical applications. Moreover, they negatively affect the classification performances of the SVM. For predicting the rarest objects, if the training instances of the majority class outnumber than the other minority class, the hyperplane or decision boundary generated by SVM can be severely skewed toward the majority class, especially in the class imbalanced or overlapping data sets. Hence, this study aims to develop the robust SVM to enhance their classification performance. Moreover, another challenge is that SVM are regarded as black box analysis tools lacking of explanation capability in the classification problem. Decision boundary of SVM always lacks explicit a declarative knowledge representation since it presents a complicated mathematical pattern. Therefore, a supportive rule extraction algorithm from SVM is needed

This study aims to develop three models, that are “modified slack variables within SVM” (MS-SVM), “Genetic Algorithm based Rule Extraction Algorithm from SVM” (GASVM), and “Integration of Kernel Clustering with Genetic Algorithm based Rule Extraction Algorithm from SVM” (KCGex-SVM). The first proposed method, MS-SVM, is applied to deal with complex data. The artificial and UCI data sets are provided to evaluate to the effectiveness of MS-SVM model. By using different performance metrics, accuracy, sensitivity, and specificity, the experimental results can be compared with the original SVM demonstrating the superiority of the MS-SVM. The second and the third ones are GASVM and KCGex-SVM proposed to enhance the explanation capability of SVM and extract the rule sets. This study utilizes measurements of accuracy, coverage, fidelity, and comprehensibility to evaluate the performance of this proposed rule extraction algorithms on the UCI data sets. Results indicate that the performance of the proposed rule extraction algorithms is better than that of others Thus, the proposed rule extraction algorithms are essential analysis tools which can be effectively used in data mining fields.

A real application is introduced in this study as well. The actual medical pressure ulcer after surgical operation is employed to illustrate the superiority of our proposed MS-SVM and GASVM. The results of the real application demonstrate that our proposed methods are practical for the real world case.

Arbatli, A.D. and Akin, H.L. (1997). Rule extraction from trained neural networks using genetic algorithms. Nonlinear Analysis, 30(3), 1639-1648.
Barakat, N. and Bradley, A.P. (2007). Rule extraction from support vector machines: A sequential covering approach. IEEE Transactions on Knowledge and Data Engineering, 19(6), 729-741.
Barakat, N. and Bradley, A.P. (2010). Rule extraction from support vector machines: A review. Neurocomputing, 74(1-3), 178-190.
Barakat, N. and Diederich, J. (2004). Learning-based rule-extraction from support vector machines: Performance on benchmark data sets. Proc Conf Neuro-Computing and Evolving Intelligence.
Barakat, N. and Diederich, J. (2006). Eclectic rule-extraction from support vector machines. International Journal of Computational Intelligence, 2(1), 59-62.
Bergstrom, N. (2005) Patients at risk for pressure ulcers and evidence-based care for pressure ulcer prevention. Pressure ulcer research: Current and future perspectives. Springer Verlang, Berlin, Germany.
Bergstrom, N., Braden, B., Laguzza, A., and Holman, V. (1987). The braden scale for predicting pressure sore risk. Nursing Research, 36(4), 205-210.
Braden, B. and Bergstrom, N. (1987). A conceptual schema for the study of the etiology of pressure sores. Rehabilitation nursing, 12(1), 8-12.
Breiman, L., Friedman, J., Olshen, R., and Stone, C. (1994) Classification and regression trees. Wadsworth and Brooks, Monterey, CA.
Burges, C.J.C. (1998). A tutorial on support vector machines for pattern recognition. Data Mining and Knowledge Discovery, 2(2), 121-167.
Chaves, A.C., Vellasco, M., and Tanscheit, R. Fuzzy rule extraction from support vector machines. In: Proceedings of the Fifth International Conference on Hybrid Intelligent Systems, 2005.
Chawla, N., Bowyer, K., Hall, L., and Kegelmeyer, W. (2002). Smote: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16, 321-357.
Cohen, W.W. (1995). Fast effective rule induction. Proc 12th Int'l Conf Maching Learning, 115-123.
Cristianini, N. and Shawe-Taylor, J. (2000) An introduction to support vector machines. Cambridge University Press, Cambridge, UK.
Davis, L.D. and Mitchell, M. (1991) Handbook of genetic algorithms. Van Nostrand Reinhold.
De Laat, E.H., Schoonhoven, L., Pickkers, P., Verbeek, A.L., and Van Achterberg, T. (2006). Implementation of a new policy results in a decrease of pressure ulcer frequency. International Journal for Quality in Health Care, 18, 107-112.
Defloor, T. (1999). The risk of pressure sores: A conceptual scheme. Journal of Clinical Nursing, 8, 206-216.
Demsar, J. (2006). Statistical comparison of classifiers over multiple data sets. Journal of Machine Learning Research, 7, 1-30.
Dhillon, I., Guan, Y., and Kulis, B. (2005). A unified view of kernel k-means, spectral clustering and graph cuts: Univ. of Texas at Austin.
Diederich, J. (2008) Rule extraction from support vector machines: An introduction. In: Rule extraction from support vector machines, vol 80. Studies in computational intelligence. Springer Berlin / Heidelberg, pp 3-31. doi:10.1007/978-3-540-75390-2_1
Frank, A. and Asuncion, A. (2010). Uci machine learning repository. (2011/02/24). Retrieved from http://archive.ics.uci.edu/ml
Fung, G., Sandilya, S., and Rao, R.B. Rule extraction from linear support vector machines. In: Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining, Chicago, Illinois, USA, 2005. ACM.
Gama, J. and Brazdil, P. Characterization of classification algorithms. In: Proceedings of the 7th Portuguese Conference on Artificial Intelligence: Progress in Artificial Intelligence, 1995. Springer-Verlag.
Garcia-Martinez, C. and Lozano, M. (2010). Evaluating a local genetic algorithm as context-independent local search operator for metaheuristics. Soft Computing, 14(10), 1117-1139.
Garcia, V., R. A. Mollineda, and Sanchez, J.S. (2008). On the k-nn performance in a challenging scenario of imbalance and overlapping. Pattern Analysis & Application, 11, 269-280.
Goldberg, D.E. (1989) Genetic algorithms in search, optimization and machine learning. Addison-Wesley Longman Publishing Co., Inc.
Gunningberg, L., Dahm, M.F., and Ehrenberg, A. (2008). Accuracy in the recording of pressure ulcers and prevention after implementing an electronic health record in hospital care. Quality and Safety in Health Care, 17(4), 281-285.
Gunningberg, L., Lindholm, C., Carlsson, M., and Sjoden, P.-O. (2001). Reduced incidence of pressure ulcers in patients with hip fractures: A 2-year follow-up of quality indicators. International Journal for Quality in Health Care, 13(5), 399-407.
Gunningberg, L. and Stotts, N.A. (2008). Tracking quality over time: What do pressure ulcer data show? International Journal for Quality in Health Care, 24, 246-253.
Guyon, I., Weston, J., Barnhill, S., and Vapnik, V. (2002). Gene selection for cancer classification using support vector machines. Machine Learning, 46(1), 389-422.
Herbrich, R. and Weston, J. Adaptive margin support vector machines for classification. In: Proceeding 9th ICANN, 1999. pp 880-885.
Ho, T.K. and Basu, M. (2002). Complexity measures of supervised classification problems. IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(3), 289-300.
Holland, J. (1975) Adaptation in natural and artificial systems. The University of Michigan Press, Ann Arbor.
Huang, K., Yang, H., King, I., and Lyu, M. Learning classifiers from imbalanced data based on biased minimax probability machine. In: Proceedings of the 04’ IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. pp 558-563.
Huang, P. and Xu, Y. (2007). Svm-based learning control of space robots in capturing operation. International Journal of Neural Systems, 17(6), 467-477.
Japkowicz, N. The class imbalance problem: Significance and strategies. In: Proceedings of the 2000 International Conference on Artificial Intelligence, 2000. pp 111-117.
Ken, B. and Sharon, A.A. (1999). Hospital-acquired pressure ulcers: A comparison of costs in medical vs. Surgical patients. Nursing Economics, 17, 263-271.
Kwong, E., Pang, S., Wong, T., Ho, J., Shao-ling, X., and Li-jun, T. (2005). Predicting pressure ulcer risk with the modified braden, braden, and norton scales in acute care hospitals in mainland china. Applied Nursing Research, 18(2), 122-128.
Lian, H.C. and Lu, B.L. (2007). Multi-view gender classification using multi-resolution local binary patterns and support vector machines. International Journal of Neural Systems, 17(6), 479-487.
Lu, W.Z. and Wang, D. (2008). Ground-level ozone prediction by support vector machine approach with a cost-sensitive classification scheme. Science of the Total Environment, 395, 109-116.
Maloof, M.A. Learning when data sets are imbalanced and when costs are unequal and unknown. In: ICML 2003 Workshop on Learning from Imbalanced Data Sets, 2003. pp 1263-1284.
Martens, D., Baesens, B., and Gestel, T.V. (2009). Decompositional rule extraction from support vector machines by active learning. IEEE Transactions on Knowledge and Data Engineering, 21(2), 178-191.
Martens, D., Baesens, B., Gestel, T.V., and Vanthienen, J. (2007). Comprehensible credit scoring models using rule extraction from support vector machines. European Journal of Operational Research, 183(3), 1466-1476.
Martens, D., Huysmans, J., Setiono, R., Vanthienen, J., and Baesens, B. (2008). Rule extraction from support vector machines: An overview of issues and application in credit scoring. Studies in Computational Intelligence, 80, 33-63.
Michalewicz, Z. (1996) Genetic algorithms + data structures = evolution programs. Springer-Verlag.
Mollineda, R.A., Sanchez, J.S., and Sotoca, J.M. (2005). Data characterization for effective prototype selection. 3523, 295-302.
Nunez, H., Angulo, C., and Catala, A. (2002). Rule extraction from support vector machines. Proc European Symp Artificial Neural Networks, 107-112.
Nunez, H., Angulo, C., and Catala, A. (2006). Rule-based learning systems for support vector machines. Neural Process Letter, 24(1), 1-18.
National Pressure Ulcer Advisory Panel (2010) Pressure ulcer category. http://www.npuap.org/resources.htm. Accessed September 1 2010.
Ortiz, M.S. (1992) Clinical trial of the braden scale for predicting pressure sore risk in a medical surgical unit. Master, University of Texas, Arlington.
Prati, R.C., Batista, G.E., and Monard, M.C. Class imbalance versus class overlapping: An analysis of a learning system behavior. In: Proceedings of the 3rd Mexican international conference on artificial intelligence, 2004. pp 312-321.
Prechelt, L. (1994). Proben1 - a set of neural network benchmark problems and benchmarking rules. Fakultat fur Informatik, Universitat Karlsruhe: Germany, Anonymous ftp.
Quinlan, J.R. (1993) Programs for machine learning. Morgan Kaufmann.
Sanchez, J., Mollineda, R., and Sotoca, J. (2007). An analysis of how training data complexity affects the nearest neighbor classifiers. Pattern Analysis & Applications, 10(3), 189-201.
Schlkopf, B. and Smola, A.J. (2002) Learning with kernels: Support vector machines, regularization, optimization, and beyond. MA: MIT Press, Cambridge.
Schoonhoven, L., Defloor, T., and Grypdonck, M.H.F. (2002a). Incidence of pressure ulcers due to surgery. Journal of Clinical Nursing, 11(4), 479-487.
Schoonhoven, L., Defloor, T., van der Tweel, I., Buskens, E., and Grypdonck, M.H.F. (2002b). Risk indicators for pressure ulcers during surgery. Applied Nursing Research, 15(3), 163-173.
Schoonhoven, L., Grobbee, D.E., Donders, A.R.T., Algra, A., Grypdonck, M.H., Bousema, M.T., Schrijvers, A.J.P., and Buskens, E. (2006). Prediction of pressure ulcer development in hospitalized patients: A tool for risk assessment. Quality and Safety in Health Care, 15, 65-70.
Setiono, R., Baesens, B., and Mues, C. (2009). A note on knowledge discovery using neural networks and its application to credit card screening. European Journal of Operational Research, 192(1), 326-332.
Sollich, P. (2002). Bayesian methods for support vector machines: Evidence and predictive class probabilities. Machine Learning, 46, 21-52.
Song, Q., Hu, W., and Xie, W. (2002). Robust support vector machine with bullet hole image classification. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, 32(4), 440-448.
Tae, Y.K. and Norma, L. Predictive modeling for the prevention of hospital-acquired pressure ulcers. In: Proc AMIA Annu Fall Symp, 2006. pp 434-438.
Tan, F., Fu, X., Zhang, Y., and Bourgeois, A. (2008). A genetic algorithm-based method for feature subset selection. Soft Computing - A Fusion of Foundations, Methodologies and Applications, 12(2), 111-120.
Tan, P.-N., Steinbach, M., and Kumar, V. (2005) Introduction to data mining. Addison Wesley.
Trafalis, T.B. and Gilbert, R.C. (2006). Robust classification and regression using support vector machines. European Journal of Operational Research, 173(3), 893-909.
Trafalis, T.B. and Gilbert, R.C. (2007). Robust support vector machines for classification and computational issues. Optimization Methods Software, 22(1), 187-198.
Trebar, M. and Steele, N. (2008). Application of distributed svm architectures in classifying forest data cover types. Computers and Electronics in Agriculture, 63, 119-130.
Vapnik, V. (1995) The nature of statistical learning theory. Springer, New York.
Veropoulos, K., Campbell, C., and Cristianini, N. Controlling the sensitivity of support vector machines. In: Proceedings of the International Joint Conference on AI,, 1999. pp 55-60.
Waterlow, J. (1985). Pressure sores: A risk assessment card. Nursing Time, 81(48), 49-55.
Willson, D., Ashton, C., Wingate, N., Goff, C., Horn, S., Davies, M., and R., B. Computerized support of pressure ulcer prevention and treatment protocols. In: Proc Annu Symp Comput Appl Med Care, 1995. pp 646-650.
Wu, G. and Chang, E.Y. (2005). Kba: Kernel boundary alignment considering imbalanced data distribution. IEEE Transactions on Knowledge and Data Engineering, 17(6), 786-794.
Zhang, J. and Mani, I. Knn approach to imbalanced data distributions: A case study involving information extraction. In: ICML 2003 Workshop on Learning from Imbalanced Data Sets, 2003.
Zielstorff, R.D., Estey, G., Vickery, A., Hamilton, G., Fitzmaurice, J.B., and Barnett, G.O. Evaluation of a decision support system for pressure ulcer prevention and management: Preliminary findings. In: Proc AMIA Annu Fall Symp, 1997. pp 248-252.