Main Content

modelDiscrimination

Compute AUROC and ROC data

Description

example

DiscMeasure = modelDiscrimination(eadModel,data) computes the area under the receiver operating characteristic curve (AUROC). modelDiscrimination supports segmentation and comparison against a reference model and alternative methods to discretize the EAD response into a binary variable.

example

[DiscMeasure,DiscData] = modelDiscrimination(___,Name,Value) specifies options using one or more name-value pair arguments in addition to the input arguments in the previous syntax.

Examples

collapse all

This example shows how to use fitEADModel to create a Tobit model and then use modelDiscrimination to compute AUROC and ROC.

Load EAD Data

Load the EAD data.

load EADData.mat
head(EADData)
ans=8×6 table
    UtilizationRate    Age     Marriage        Limit         Drawn          EAD    
    _______________    ___    ___________    __________    __________    __________

        0.24359        25     not married         44776         10907         44740
        0.96946        44     not married    2.1405e+05    2.0751e+05         40678
              0        40     married        1.6581e+05             0    1.6567e+05
        0.53242        38     not married    1.7375e+05         92506        1593.5
         0.2583        30     not married         26258        6782.5        54.175
        0.17039        54     married        1.7357e+05         29575        576.69
        0.18586        27     not married         19590          3641        998.49
        0.85372        42     not married    2.0712e+05    1.7682e+05    1.6454e+05

rng('default');
NumObs = height(EADData);
c = cvpartition(NumObs,'HoldOut',0.4);
TrainingInd = training(c);
TestInd = test(c);

Select Model Type

Select a model type for Tobit or Regression.

ModelType = "Tobit";

Select Conversion Measure

Select a conversion measure for the EAD response values.

ConversionMeasure = "LCF";

Create Tobit EAD Model

Use fitEADModel to create a Tobit model using EADData.

eadModel = fitEADModel(EADData,ModelType,'PredictorVars',{'UtilizationRate','Age','Marriage'}, ...
    'ConversionMeasure',ConversionMeasure,'DrawnVar','Drawn','LimitVar','Limit','ResponseVar','EAD');
disp(eadModel);
  Tobit with properties:

        CensoringSide: "both"
            LeftLimit: 0
           RightLimit: 1
              ModelID: "Tobit"
          Description: ""
      UnderlyingModel: [1x1 risk.internal.credit.TobitModel]
        PredictorVars: ["UtilizationRate"    "Age"    "Marriage"]
          ResponseVar: "EAD"
             LimitVar: "Limit"
             DrawnVar: "Drawn"
    ConversionMeasure: "lcf"

Display the underlying model. The underlying model's response variable is the transformation of the EAD response data. Use the 'LiimitVar' and 'DrwanVar' name-value arguments to modify the transformation.

disp(eadModel.UnderlyingModel);
Tobit regression model:
     EAD_lcf = max(0,min(Y*,1))
     Y* ~ 1 + UtilizationRate + Age + Marriage

Estimated coefficients:
                             Estimate         SE         tStat       pValue 
                            __________    __________    ________    ________

    (Intercept)                0.22735      0.025045      9.0776           0
    UtilizationRate            0.47364      0.016536      28.643           0
    Age                     -0.0013929    0.00061488     -2.2654    0.023537
    Marriage_not married     -0.006888      0.012113    -0.56863     0.56964
    (Sigma)                    0.36419     0.0038746      93.995           0

Number of observations: 4378
Number of left-censored observations: 0
Number of uncensored observations: 4377
Number of right-censored observations: 1
Log-likelihood: -1791.06

Predict EAD

EAD prediction operates on the underlying compact statistical model and then transforms the predicted values back to the EAD scale. You can specify the predict function with different options for the 'ModelLevel' name-value argument.

predictedEAD = predict(eadModel,EADData(TestInd,:),'ModelLevel','ead');
predictedConversion = predict(eadModel,EADData(TestInd,:),'ModelLevel','ConversionMeasure');

Validate EAD Model

For model validation, use modelDiscrimination, modelDiscriminationPlot, modelAccuracy, and modelAccuracyPlot.

Use modelDiscrimination and then modelDiscriminationPlot to plot the ROC curve.

ModelLevel = "ConversionMeasure";

[DiscMeasure1,DiscData1] = modelDiscrimination(eadModel,EADData(TestInd,:),'ModelLevel',ModelLevel)
DiscMeasure1=table
              AUROC 
             _______

    Tobit    0.70892

DiscData1=1534×3 table
        X             Y           T   
    __________    _________    _______

             0            0     0.6399
             0    0.0027778     0.6399
             0    0.0041667     0.6388
    0.00096993    0.0055556    0.63771
    0.00096993    0.0069444    0.63662
    0.00096993    0.0083333    0.63558
     0.0019399    0.0097222    0.63552
     0.0019399       0.0125    0.63448
     0.0029098     0.013889    0.63442
     0.0029098     0.020833    0.63339
     0.0029098     0.026389    0.63333
     0.0029098     0.027778    0.63317
     0.0038797     0.027778      0.633
     0.0038797     0.029167    0.63259
     0.0058196     0.029167    0.63229
     0.0067895     0.031944    0.63223
      ⋮

modelDiscriminationPlot(eadModel,EADData(TestInd, :),'ModelLevel',ModelLevel,'SegmentBy','Marriage');

Figure contains an axes object. The axes object with title E A D _ l c f blank R O C blank S e g m e n t e d blank b y blank M a r r i a g e contains 2 objects of type line. These objects represent Tobit, married, AUROC = 0.70789, Tobit, not married, AUROC = 0.70898.

Input Arguments

collapse all

Exposure at default model, specified as a previously created Regression or Tobit object using fitEADModel.

Data Types: object

Data, specified as a NumRows-by-NumCols table with predictor and response values. The variable names and data types must be consistent with the underlying model.

Data Types: table

Name-Value Arguments

Specify optional comma-separated pairs of Name,Value arguments. Name is the argument name and Value is the corresponding value. Name must appear inside quotes. You can specify several name and value pair arguments in any order as Name1,Value1,...,NameN,ValueN.

Example: [DiscMeasure,DiscData] = modelDiscrimination(eadModel,data(TestInd,:),'DataID','Testing','DiscretizeBy','median')

Data set identifier, specified as the comma-separated pair consisting of 'DataID' and a character vector or string. The DataID is included in the output for reporting purposes.

Data Types: char | string

Discretization method for EAD data at the defined ModelLevel, specified as the comma-separated pair consisting of 'DiscretizeBy' and a character vector or string.

  • 'mean' — Discretized response is 1 if observed EAD is greater than or equal to the mean EAD, 0 otherwise.

  • 'median' — Discretized response is 1 if observed EAD is greater than or equal to the median EAD, 0 otherwise.

Data Types: char | string

Name of a column in the data input, not necessarily a model variable, to be used to segment the data set, specified as the comma-separated pair consisting of 'SegmentBy' and a character vector or string. One AUROC is reported for each segment, and the corresponding ROC data for each segment is returned in the optional output.

Data Types: char | string

Model level, specified as the comma-separated pair consisting of 'ModelLevel' and a character vector or string.

Note

Regression models support all three model levels, but a Tobit model supports only a ModelLevel for 'ead' and 'conversionMeasure'.

Data Types: char | string

EAD values predicted for data by the reference model, specified as the comma-separated pair consisting of 'ReferenceEAD' and a NumRows-by-1 numeric vector. The modelDiscrimination output information is reported for both the eadModel object and the reference model.

Data Types: double

Identifier for the reference model, specified as the comma-separated pair consisting of 'ReferenceID' and a character vector or string. 'ReferenceID' is used in the modelDiscrimination output for reporting purposes.

Data Types: char | string

Output Arguments

collapse all

AUROC information for each model and each segment, returned as a table. DiscMeasure has a single column named 'AUROC' and the number of rows depends on the number of segments and whether you use a ReferenceID for a reference model . The row names of DiscMeasure report the model IDs, segment, and data ID.

ROC data for each model and each segment, returned as a table. There are three columns for the ROC data, with column names 'X', 'Y', and 'T', where the first two are the X and Y coordinates of the ROC curve, and T contains the corresponding thresholds. For more information, see Model Discrimination or perfcurve.

If you use SegmentBy, the function stacks the ROC data for all segments and DiscData has a column with the segmentation values to indicate where each segment starts and ends.

If reference model data is given, the DiscData outputs for the main and reference models are stacked, with an extra column 'ModelID' indicating where each model starts and ends.

More About

collapse all

Model Discrimination

Model discrimination measures the risk ranking.

The modelDiscrimination function computes the area under the receiver operator characteristic (AUROC) curve, sometimes called simply the area under the curve (AUC). This metric is between 0 and 1 and higher values indicate better discrimination.

To compute the AUROC, you need a numeric prediction and a binary response. For EAD models, the predicted EAD is used directly as the prediction. However, the observed EAD must be discretized into a binary variable. By default, observed EAD values greater than or equal to the mean observed EAD are assigned a value of 1, and values below the mean are assigned a value of 0. This discretized response is interpreted as “high EAD” vs. “low EAD.” Therefore, the modelDiscrimination function measures how well the predicted EAD separates the “high EAD” vs. the “low EAD” observations. You can change the level to compute the model discrimination with the ModelLevel name-value pair argument and the discretization criterion with the DiscretizeBy name-value pair argument.

To plot the receiver operator characteristic (ROC) curve, use the modelDiscriminationPlot function. However, if you need the ROC curve data, use the optional DiscData output argument from the modelDiscrimination function.

The ROC curve is a parametric curve that plots the proportion of

  • High EAD cases with predicted EAD greater than or equal to a parameter t, or true positive rate (TPR)

  • Low EAD cases with predicted EAD greater than or equal to the same parameter t, or false positive rate (FPR)

The parameter t sweeps through all the observed predicted EAD values for the given data. The DiscData optional output contains the TPR in the 'X' column, the FPR in the 'Y' column, and the corresponding parameters t in the 'T' column. For more information about ROC curves, see Performance Curves.

References

[1] Baesens, Bart, Daniel Roesch, and Harald Scheule. Credit Risk Analytics: Measurement Techniques, Applications, and Examples in SAS. Wiley, 2016.

[2] Bellini, Tiziano. IFRS 9 and CECL Credit Risk Modelling and Validation: A Practical Guide with Examples Worked in R and SAS. San Diego, CA: Elsevier, 2019.

[3] Brown, Iain. Developing Credit Risk Models Using SAS Enterprise Miner and SAS/STAT: Theory and Applications. SAS Institute, 2014.

[4] Roesch, Daniel and Harald Scheule. Deep Credit Risk. Independently published, 2020.

Introduced in R2021b