Definition
Accuracy = (TP + TN) / (TP + FP + FN + TN).
With balanced data (π ≈ 0.5), Accuracy ≈ 0.5·TPR + 0.5·TNR.
TPR0.70
TNR0.80
Accuracy0.75
Confusion Matrix (N = 100, π = 0.5)
Pred +
Pred −
Actual +
Actual −
Move the sliders—or hit Play. The point is at (TNR on x, TPR on y). Accuracy is highest near the top-right corner (both high).