Calculator Inputs
Formula Used
Percentage: FNR × 100
Recall relationship: FNR = 1 − Recall
False negatives are missed actual positives. True positives are correctly detected positives. Lower FNR usually indicates better positive-case detection.
How to Use
- Select binary, multiclass, model comparison, or threshold mode.
- Enter non-negative confusion matrix counts or CSV-style rows.
- Choose decimal precision and optional false-negative cost.
- Submit the form to calculate results and interpretation.
- Copy, print, or export the result as CSV or PDF.
Example Data
| Scenario | TP | FN | Calculation | FNR |
|---|---|---|---|---|
| Medical screening | 85 | 15 | 15 ÷ 100 | 15% |
| Fraud detection | 190 | 10 | 10 ÷ 200 | 5% |
| Defect inspection | 72 | 28 | 28 ÷ 100 | 28% |
Practical Guidance
Medical screening often prioritises very low false negative rates. Fraud systems may balance missed fraud against investigation workload. Safety systems usually require conservative threshold validation.
Never judge a classifier using FNR alone. Review precision, FPR, calibration, and prevalence together. Validate decisions with representative production data.
Frequently Asked Questions
What does false negative rate measure?
It measures missed positives among all actual positive cases.
Is FNR the same as miss rate?
Yes. False negative rate is commonly called the miss rate.
How is FNR related to recall?
FNR equals one minus recall. Their percentages always total 100%.
Can FNR exceed 100%?
No. Valid FNR values range from zero through one.
What happens when TP and FN equal zero?
The denominator becomes zero. FNR is therefore undefined.
Is a lower FNR always better?
Usually, but reducing FNR can increase false positives significantly.
How should multiclass FNR be reported?
Report per-class results plus macro, micro, and weighted averages.
Why include false-negative cost?
Costs translate missed cases into an operational business estimate.
Which threshold should I choose?
Choose using validation data, costs, prevalence, and capacity limits.
Why can small samples be misleading?
One changed prediction can strongly alter rates in small datasets.