Download Understanding Confusion Matrix: TP, TN, FP, FN and more Study notes Statistics in PDF only on Docsity!
Statistics calculated on confusion matrix
Confusion matrix
Theoretical confusion matrix
Automatic classification Positive Negative Manual classification Positive TP FN Negative FP TN
Example
Automatic classification of three groups A , B , C Automatic classification A B C Manual classification
A 1 1 1
B 1 3 0
C 0 0 3
For group A , the matrix can be reduced as: Automatic classification Group A Not group A Manual classification Group A TP = 1 FN = 2 Not group A FP = 1 TN = 6 For group B , the matrix can be reduced as: Automatic classification Group B Not group B Manual classification Group B TP = 3 FN = 1 Not group B FP = 1 TN = 5 For group C , the matrix can be reduced as: Automatic classification Group C Not group C Manual classification Group C TP = 3 FN = 0 Not group C FP = 1 TN = 6
Explanation of main parameters
TP : True Positive
Number of particles of the group of interest correctly classified.
TN : True Negative
Number of particles of all the other groups classified as other groups.
FP : False Positive
Number of particles of other groups classified in the group of interest.
FN : False Negative
Number of particles of the group of interest classified in the other groups.
General statistics
Accuracy
( TP + TN )
( TP + TN + FP + FN )
Error
1 – Accuracy
FOR : False Omission Rate
FN
( FN + TN )
= 1 – NPV
The four ratios of ratios
LRPT : Likelihood Ratio for Positive Tests
TP /( TP + FN )
FP /( FP + TN )
Recall ( 1 − Specificity )
Recall FPR
LRNT : Likelihood Ratio for Negative Tests
FN /( FN + TP )
TN /( TN + FP )
( 1 − Recall ) Specificity
FNR
Specificity
LRPS : Likelihood Ratio for Positive Subjects
TP /( TP + FP )
FN /( FN + TN )
Precision ( 1 − NPV )
Precision FOR
LRNS : Likelihood Ratio for Negative Subjects
FP /( FP + TP )
TN /( TN + FN )
( 1 − Precision ) ( 1 − FOR )
FDR
NPV
