對應用統計方法評估核函數分類能力的研究(編輯修改稿)

　

【文章內容簡介】 )31 (3)6 (6)87 (9)（表中數據代表核函數兩兩對比的獲勝次數與失敗次數之差，括號內的數字指明該核函數在當前評估準則下的排名）從表7的統計結果中可以得出以下結論：(1) 雖然根據核函數兩兩對比的獲勝次數與失敗次數之差對核函數分類能力進行排序時，不同評估準則在具體數值上存在差異，但應用統計方法所獲得的核函數排序是大體一致的；(2) 對核函數分類能力的評估結論與實驗1中結論一致.5討論核函數選擇的準則和方法作為核方法及其應用的核心內容之一，目前在國際上還沒有形成一個統一的模式，在解決實際問題時，人們往往只能是憑借經驗，并采用試湊方式，.本文嘗試將糾正重復取樣 t 測試的統計方法應用到核函數選擇中，，不同統計方法之間也存在差異，判斷核函數分類能力要在應用多種統計方法的基礎上，綜合考評準確率、精確率、召回率、F 測量和信息增益等多種評估準則，真正好的核函數分類模型在所有評估準則下所獲得的結果都應當是優良的， 統計方法是計算密集型的方法，為了縮短運算時間，可以考慮數據壓縮以減少參加運算的數據量.第 27 頁 共 27 頁表5 實驗中使用的核函數[ 8 ]英文名稱縮寫數學表達式參數定義參考函數Linear FunctionLinearPolynomial FunctionPF,default 3Radial Basis Function or Gaussian FunctionRBF,default 1Symmetric Triangle FunctionSTF,default 3Cauchy FunctionCF,default 3Laplace FunctionLF,default 3Hyperbolic Secant FunctionHSF,default 3Squared Sin Cardinal Function or Squared Sinc FunctionSSCF,default 3Hyperbolic Tangent Function or Sigmoid FunctionHTF,default 1參考文獻[1] Vapnik V.. The nature of statistical learning theory, Second Edition, New York: SpringerVerlag, 2000.[2] Steinwart I.. On the influence of the kernel on the consistency of support vector machines. Journal of Machine Learning Research, 2002, 2: 6793.[3] Chalimourda A., Sch246。lkopf B., Smola A.. Experimentally optimal v in support vector regression for different noise models and parameter settings. Neural Networks, 2004, 17(1): 127141.[4] Liu ., Luo B., Chen .. Optimal model selection for support vector machines. Journal of Computer Research and Development, 2005, 42(4): 576581 (in Chinese)(劉向東, 駱斌, 陳兆乾. 支持向量機最優模型選擇的研究. 計算機研究與發展, 2005, 42(4): 576581)[5] Wang L., Bo ., Liu F., Jiao .. Least squares hidden space support vector machines. Chinese Journal of Computers, 2005, 28(8): 13021307 (in Chinese)(王玲, 薄列峰, 劉芳, 焦李成. 最小二乘隱空間支持向量機. 計算機學報, 2005, 28(8): 13021307)[6] Wu T., He ., He .. Interpolation based kernel function’s construction. Chinese Journal of Computers, 2003, 26(8): 990996 (in Chinese)(吳濤, 賀漢根, 賀明科. 基于插值的核函數構造. 計算機學報, 2003, 26(8): 990996)[7] Tan Y., Wang J.. A support vector machine with a hybrid kernel and minimal vapnikchervonenkis dimension. IEEE Trans. on Knowledge and Data Engineering, 2004, 16(4): 385395.[8] Chen ., Wang .. Support vector learning for fuzzy rulebased classification systems. IEEE Trans. on Fuzzy System, 2003, 11(6): 716728.[9] Browne .. Crossvalidation methods. Journal of Mathematical Psychology, 2000, 44: 108132.[10] Sincich T.. Business statistics by example, Fifth Edition. New Jersey: Prentice Hall, 1996.[11] Nadeau C., Bengio Y.. Inference for the generalization error. Machine Learning, 2003, 52(3): 239281.[12] Cover .. Elements of information theory, Second Edition. New Jersey: John Wiley amp。 Sons, 2006.[13] Racine J.. Consistent crossvalidatory modelselection for dependent data: hvblock crossvalidation. Journal of Econometrics, 2000, 99(1): 3961. [14] Witten ., Frank E.. Data mining: practical machine learning tools and techniques, Second Edition. San Francisco: Morgan Kaufmann, 2005.(董琳 等譯. 數據挖掘－實用機器學習技術(第二版). 機械工業出版社, 2006)[15] Yong Wang, Hu .. Study of the relationship between normalized information gain and accuracy、precision and recall. Beijing, 2007 Chinese Conference on Pattern Recognition (CCPR 2007), Beijing, Science Press, 2007: 2734 (in Chinese)(王泳, 胡包鋼. 歸一化信息增益準則與準確率、精確率、召回率的非線性關系研究. 2007年全國模式識別學術會議（CCPR2007）,北京,2007年12月,科學出版社, 2007: 2734)[16] Hu ., Yong Wang. Applications of Mutual Information Criteria in Classification Problems. Beijing, 2007 Chinese Conference on Pattern Recognition (CCPR 2007), Beijing, Science Press, 2007: 3545 (in Chinese)(胡包鋼, 王泳. 關于互信息學習準則在分類問題中的應用. 2007年全國模式識別學術會議（CCPR2007）,北京, 2007年12月, 科學出版社, 2007: 3545)[17] Newman ., Hettich S., Blake ., Merz .. UCI repository of machine learning databases [~mlearn/]. Irvine, CA: University of California, Department of Information and Computer Science, 1998.[18] Statlib – Data, Software and News from the Statistics Community. [][19] Cauwenberghs G., Poggio T.. Incremental and decremental support vector machine learning. Advances in