The ROC-curve

Credit Risk Modeling in R

Lore Dirick

Manager of Data Science Curriculum at Flatiron School

Until now

Strategy table/curve : still make assumption
What is "overall" best model?

Confusion matrix

$$

Actual loan status v. Model prediction

	No default (0)	Default (1)
No default (0)	TN	FP
Default (1)	FN	TP

$$

$\text{Accuracy} = \frac{TP +TN}{TP + FP + TN + FN}$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

Accuracy?

$$

Screen Shot 2020-06-22 at 6.21.14 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

The ROC-curve

Screen Shot 2020-06-22 at 6.21.58 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

The ROC-curve

Screen Shot 2020-06-22 at 6.22.11 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

The ROC-curve

Screen Shot 2020-06-22 at 6.22.22 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

The ROC-curve

Screen Shot 2020-06-22 at 6.22.35 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

The ROC-curve

Screen Shot 2020-06-22 at 6.22.45 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

The ROC-curve

Screen Shot 2020-06-22 at 6.22.54 PM.png

$$

$\text{Sensitivity} = \frac{TP}{TP + FN}$

$\text{Specificity} = \frac{TN}{TN + FP}$

Which one is better?

Screen Shot 2020-06-22 at 6.23.06 PM.png

AUC ROC-curve A = 0.75
AUC ROC-curve B = 0.78

Let's practice!

Credit Risk Modeling in R

Preparing Video For Download...