비용곡선과 ROC곡선에서의 비용비율 = Cost Ratios for Cost and ROC Curves

다국어 초록 (Multilingual Abstract)

For classification problems on mixture distribution, a threshold based on cost functions is optimal from the viewpoint of a minimum expected cost. Assuming that there is no cost information, we propose cost ratios in the expected cost corresponding to thresholds where the total accuracy and the true rate are maximized to explain the relation of these cost ratios minimizing the expected cost. Other cost ratios are also proposed by comparing the normalized expected costs when classification accuracy is maximized. The values of these cost ratios are located between two cost ratios for the expected costs based on classification accuracies, and converge to that of the minimum expected cost. This work suggests two cost ratios: one is minimized by the expected cost and the normalized expected cost, and the other in the expected cost and the normalized expected cost functions that are maximized classification accuracies. We discuss their compatibility based on the relation of these cost ratios.

참고문헌 (Reference)

1 홍종선, "혼합분포에서 최적분류점" 한국통계학회 23 (23): 13-28, 2010

2 Tasche, D., "Validation of Internal Rating Systems and PD Estimates"

3 Briggs, W. M., "The skill plot: a graphical technique for the evaluating the predictive usefulness of continuous diagnostic tests" 63 : 250-261, 2007

4 Davis, J., "The relationship between precision-recall and ROC curves" 233-240, 2006

5 Liu, Y., "The evaluation of classification models for credit scoring" 2 : 1-65, 2002

6 Pepe, M. S., "The Statistical Evaluation of Medical Tests for Classification and Prediction" University Press 2003

7 Zhou, X. H., "Statistical Methods in Diagnostic Medicine" Wiley 2002

8 Hilden, J., "Regret graphs, diagnostic uncertainty and Youden’s index" 15 : 969-986, 1996

9 Zweig, M. H., "Receiver-operating characteristic(ROC) plots: A fundamental evaluation tool in clinical medicine" 39 : 561-577, 1993

10 홍종선, "ROC와 CAP 곡선에서의 최적 분류점" 한국통계학회 22 (22): 911-921, 2009

11 Fawcett, T., "ROC graphs with instance-varying costs" 27 : 882-891, 2006

12 Vuk, M., "ROC curve, lift chart and calibration plot" 3 : 89-108, 2006

13 김지현, "ROC and Cost Graphs for General Cost Matrix Where Correct Classifications Incur Non-zero Costs" 한국통계학회 11 (11): 21-30, 2004

14 Krzanowski, W. J., "ROC Curves for Continuous Data" Champman & Hall/CRC 2009

15 Hand, D. J., "Mismatched models, wrong results, and dreadful decisions: on choosing appropriate data mining tools" 2009

16 Jund, J., "Methods to estimate the optimal threshold for normally or log-narmally distributed biological tests" 25 : 406-415, 2005

17 Servigny, A. D., "Measuring and Managing Credit Risk" McGraw-Hill 2004

18 Kaivanto, K., "Maximization of the sum of sensitivity and specificity as a diagnostic cutpoint criterion" 61 : 516-518, 2008

19 Antonie, M. L., "Learning to use a learned model: A two-stage approach to classication" 33-42, 2006

20 Hand, D. J., "Evaluating models for classifying customers in retail banking collections" 2009

21 Holte, R. C., "Cost-sensitive classifier evaluation using cost curves" 5012 : 26-29, 2008

22 Turney, P. D., "Cost-sensitive classification: Empirical evaluation of a hybrid genetic decision tree induction algorithm" 2 : 369-409, 1995

23 Drummond, C., "Cost curves: an improved method for visualizing classifier performance" 65 : 95-130, 2006

24 Liu, Y., "Comparing evaluation metrics for sentence boundary detection" 185-188, 2007

25 Adams, N. M., "Comparing classifiers when the misallocation costs are uncertain" 30 : 1139-1147, 1999

26 Metz, C. E., "Basic principles of ROC analysis" 8 : 283-298, 1978

27 Hoshino, R., "Applications and extensions of cost curves to marine container inspection" 2009

28 Provost, F., "Analysis and visualization of classifier performance: comparison under imprecise class and cost distributions" 43-48, 1997

29 Fielding, A. H., "A review of methods for the measurement of prediction errors in conservation presence/absence models" 24 : 38-49, 1997

30 Cantor, S. B., "A comparison of C/B ratios from studies using receiver operating characteristic curve analysis" 52 : 885-892, 1999

31 Velez, D. R., "A balanced accuracy function for epistasis modeling in imbalanced datasets using multifactor dimensionality reduction" 31 : 306-315, 2007