• The Korean Journal of Applied Statistics
• /
• v.24 no.2
• /
• pp.269-278
• /
• 2011

There are many researches that have considered the distribution functions and appropriate covariates corresponding to the scores in order to improve the accuracy of a diagnostic test, including the ROC curve that is represented with the relations of the sensitivity and the specificity. The ROC analysis was used by the regression model including some covariates under the assumptions that its distribution function is known or estimable. In this work, we consider a general situation that both the distribution function and the elects of covariates are unknown. For the ROC analysis, the mixtures of normal distributions are used to estimate the distribution function fitted to the credit evaluation data that is consisted of the score random variable and two sub-populations of parameters. The AUC measure is explored to compare with the nonparametric and empirical ROC curve. We conclude that the method using normal mixtures is fitted to the classical one better than other methods.

• The Korean Journal of Applied Statistics
• /
• v.24 no.6
• /
• pp.987-994
• /
• 2011

From the point view of credit evaluation whose population is divided into the default and non-default state, two methods are considered to estimate conditional distribution functions: one is to estimate under the assumption that the data is followed the mixture normal distribution and the other is to use the kernel density estimation. The parameters of normal mixture are estimated using the EM algorithm. For the kernel density estimation, five kinds of well known kernel functions and four kinds of the bandwidths are explored. In addition, the corresponding ROC functions are obtained based on the estimated distribution functions. The goodness-of-fit of the estimated distribution functions are discussed and the performance of the ROC functions are compared. In this work, it is found that the kernel distribution functions shows better fit, and the ROC function obtained under the assumption of normal mixture shows better performance.

• Communications for Statistical Applications and Methods
• /
• v.19 no.2
• /
• pp.277-286
• /
• 2012

For credit assessment models, the ROC curves evaluate the classification performance using two univariate cumulative distribution functions of the false positive rate and true positive rate. In this paper, it is extended to two bivariate normal distribution functions of default and non-default borrowers; in addition, the bivariate ROC curves are proposed to represent the joint cumulative distribution functions by making use of the linear function that passes though the mean vectors of two score random variables. We explore the classification performance based on these ROC curves obtained from various bivariate normal distributions, and analyze with the corresponding AUROC. The optimal threshold could be derived from the bivariate ROC curve using many well known classification criteria and it is possible to establish an optimal cut-off criteria of bivariate mixture distribution functions.

• The Korean Journal of Applied Statistics
• /
• v.33 no.5
• /
• pp.541-553
• /
• 2020

The receiver operating characteristic (ROC) curve is expressed as both sensitivity and specificity; in addition, some optimal thresholds using the ROC curve are also represented with both sensitivity and specificity. In addition to the sensitivity and specificity, the expected usefulness function is considered as disease prevalence and usefulness. In particular, partial the area under the ROC curve (AUC) on a certain range should be compared when the AUCs of the crossing ROC curves have similar values. In this study, partial AUCs representing high sensitivity and specificity are proposed by using sensitivity and specificity lines, respectively. Assume various distribution functions with ROC curves that are crossing and AUCs that have the same value. We propose a method to improve the discriminant power of the classification models while comparing the partial AUCs obtained using sensitivity and specificity lines.

• The Korean Journal of Applied Statistics
• /
• v.32 no.4
• /
• pp.593-605
• /
• 2019

Significant literature exists on the area under the ROC curve (AUC) and the volume under the ROC surface (VUS) which are statistical measures of the discriminant power of classification models. Whereas the partial AUC is restricted on the false positive rate, the two-way partial AUC is restricted on both the false positive rate and true positive rate, which could be more efficient and accurate than partial AUC. The two-way partial AUC was suggested as more efficient and accurate than the partial AUC. Partial VUS as well as the three-way partial VUS were also developed for the ROC surface. A proposed AUC is expressed in this paper with probability and integration using two truncated distribution functions restricted on both the false positive rate and true positive rate. It is also found that this AUC has a relation with the two-way partial AUC. The three-way partial VUS for the ROC surface is also related to the VUS using truncated distribution functions. These AUC and VUS are represented and estimated in terms of Mann-Whitney statistics. Their parametric and non-parametric estimation methods are explored based on normal distributions and random samples.

• The Korean Journal of Applied Statistics
• /
• v.32 no.2
• /
• pp.187-198
• /
• 2019

Extensive literature exists on how to estimate optimal thresholds based on various accuracy measures using receiver operating characteristic (ROC) and cumulative accuracy profile (CAP) curves. This paper now proposes an alternative measure to represented the specific partial area under the ROC and CAP curves. The relationship between ROC and CAP functions is examined using differential equations of the new defined partial area under curves. In addition, the relationship with the optimal thresholds under conditions of various accuracy measures for the ROC and CAP functions is also derived. We assume there are two kinds of distribution functions composing the mixed distribution as various normal distributions before finding the optimal thresholds. Corresponding type 1 and 2 errors are also explored and discussed under various conditions for accuracy measures.

• The Korean Journal of Applied Statistics
• /
• v.22 no.1
• /
• pp.29-39
• /
• 2009

Among others, ROC and CAP curves are used to explore the discriminatory power between the defaults and non-defaults, based on the distribution of the probability of default in credit rating works. ROC and CAP curves are plotted in terms of various ratios of the probability of default. Each point on ROC and CAP curves is calculated according to cutting points (scores) for classifying between defaults and non-defaults. In this paper, adjusted ROC and CAP curves are proposed by using functions of ratios of the probability of default. It is possible to recognize the score corresponding to a point oil these adjusted curves, and we can identify the best score to show the optimal discriminatory power. Moreover, we discuss the relationships between the best score obtained from the adjusted ROC and CAP curves and the score corresponding to Kolmogorov - Smirnov statistic to test the homogeneous distribution functions of the defaults and non-defaults.

• The Korean Journal of Applied Statistics
• /
• v.22 no.5
• /
• pp.911-921
• /
• 2009

Receiver Operating Characteristic(ROC) and Cumulative Accuracy Profile(CAP) curves are two methods used to assess the discriminatory power of different credit-rating approaches. The points of optimal classification accuracy on an ROC curve and of maximal profit on a CAP curve can be found by using iso-performance tangent lines, which are based on the standard notion of accuracy. In this paper, we offer an alternative accuracy measure called the true rate. Using this rate, one can obtain alternative optimal threshold points on both ROC and CAP curves. For most real populations of borrowers, the number of the defaults is much less than that of the non-defaults, and in such cases the true rate may be more efficient than the accuracy rate in terms of cost functions. Moreover, it is shown that both alternative scores of optimal classification accuracy and maximal profit are the identical, and this single score coincides with the score corresponding to Kolmogorov-Smirnov statistic used to test the homogeneous distribution functions of the defaults and non-defaults.

• The Korean Journal of Applied Statistics
• /
• v.27 no.5
• /
• pp.773-786
• /
• 2014

We propose the multiplication of false rates (MFR) which is a classification accuracy criteria and an area type of rectangle from ROC curve. Optimal threshold obtained using MFR is compared with other criteria in terms of classification performance. Their optimal thresholds for various distribution functions are also found; consequently, some properties and advantages of MFR are discussed by comparing FNR and FPR corresponding to optimal thresholds. Based on general cost function, cost ratios of optimal thresholds are computed using various classification criteria. The cost ratios for cost curves are observed so that the advantages of MFR are explored. Furthermore, the de nition of MFR is extended to multi-dimensional ROC analysis and the relations of classification criteria are also discussed.

• Journal of the Korean Data and Information Science Society
• /
• v.25 no.3
• /
• pp.473-483
• /
• 2014

Even though the ROC manifold for more than three dimensional space which is an extension of the ROC curve and surface has difficulty to represent graphically, the hypervolume under the ROC manifold (HUM) statistic can be defined and obtained based on AUC and VUS measures for the ROC curve and the ROC surface. Hence the definition and characteristics of the HUM for four dimensional space are studied in this work. By extension of the standard criterion of AUC for probabilities of default based on Basel II, the 13 classes of standard criterion of HUM are proposed in order to discriminate four classification models and some application methods are discussed. In order to explore the standard criterion of HUM whose values are obtained from various distributions, ternary plot is used and explained.