ROC curve

Confusion matrix example

	Predicted A=0	Predicted A=1
Fact A=0	25	10
Fact A =1	5	30

Confusion Matrix definition

	Predicted A=0 (negative)	Predicted A=1 (positive)
Fact A=0 (negative)	a	b
Fact A =1 (positive)	c	d

Further definitions

	Predicted A=0 (negative)	Predicted A=1 (positive)
Fact A=0 (negative)	TN=a/(a+b)	FP=b/(a+b)
Fact A =1 (positive)	FN=c/(c+d)	TP=recall =d/(c+d)
		Precision P=d/(b+d)	Accuracy AC=(a+d)/(a+b+c+d)

ROC graph: x=FP and y=TP

You can also graph: x=P and y=AC and other pairs from the last table.

In addition you can read 1 and 2. Note that the notation differs from used above.

Table: Schematic outcomes of a test.

More terms

	Predicted A=0 (negative)	Predicted A=1 (positive)
Fact A=0 (negative)	TN=specificity= a/(a+b)	FP=(1-specificity)= b/(a+b)
Fact A =1 (positive)	FN=(1-sencitivity) = c/(c+d)	TP=recall=sensitivity =d/(c+d)

Accuracy =AC=(a+d)/(a+b+c+d)

Positive Likelihood Ratio= TP/FP

Negative Likelihood Ratio= FN/TN

Positive Predictive Value=Precision=P= d/(d+b)

Negative Predictive Value = a/(c+a)

From 1