Hello everyone
Various articles that study image processing of machine learning algorithms, use different outcomes for measurements. I want to know that can we convert any of these measurements to each other, in order to do a meta-analysis of them?
Accuracy (ACC)
Precision (PRE)
F1 score
Sensitivity
Specificity
Recall (REC)
Area Under Curve (AUC)
Hey Mahmood Dashti,
That is a really great question.
These forms of measurements and calculations are based on one common fundamental entity called Confusion Matrix which is again calculated on four of these values:
True Positive (TP)
True Negative (TN)
False Positive (FP)
False Negative (FN)
Precision = True positives/ (True positives + False positives) = TP/(TP + FP)
Accuracy = (True positives + True Negatives)/ (True positives + True negatives + False positives + False negatives)
= (TP + TN)/ (TP + TN + FP + FN)
Recall = True positives/ (True positives + False negatives)
= TP/ (TP + FN)
The above three are again used to calc. F1 score which is:
F1 = 2 * (precision * recall) / (precision + recall)
The AUC-ROC Curve again is again the graph plotted for TPR (True Positive Rate)/Sensitivity/Recall vs FPR (False Positive Rate)
where again
TPR/Recall/Sensitivity = TP/(TP+FN)
Specificity = TN/(TN+FP)
FPR = 1- Specificity = FP/(TN+FP)
All these metrics are used to observe and analyze the data accordingly. The common point of evaluation stays the same.
Dear Rahul Pandya
Thank you for explaining these so well.
So Recall and sensitivity are the same thing?
And for F1 score, we use percentage for precision and recall scores, we use that percentage values in the equation?
Hey Mahmood Dashti
Recall and Sensitivity is the same. It is the number of positive records correctly predicted.
While solving the equations implement the coefficient values.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis