'Confusion matrix' is:

Confusion Matrix for binary classification tabulates TP (correctly predicted positive — fraud correctly flagged), TN (correctly predicted negative — legitimate transaction correctly cleared), FP (false alarm — legitimate flagged as fraud, Type I error), FN (missed detection — fraud not flagged, Type II error).  Metrics based on Confusion Matrix: • Accuracy = (TP+TN)/(TP+TN+FP+FN) • Precision = TP/(TP+FP) — of flagged frauds, how many are real?  • Recall/Sensitivity = TP/(TP+FN) — of all frauds, how many did we catch?  • F1 = 2×(Precision×Recall)/(Precision+Recall).  In banking fraud detection, high Recall is critical as missing a fraud (FN) is more costly than false alarms (FP).