Calculators
Mean Absolute Percentage Error Calculator
Calculate MAPE from paired actual and predicted values privately in your browser, with clear validation for zero actual values.
Measure prediction errors as percentages of actual values.
MAPE = (100 / n) × Σ| (predicted − actual) / actual |. Each pair contributes equally to the average percentage error. For actual values 100, 200, 50 and predictions 90, 220, 50, MAPE is 6.66666666667%. A perfect prediction has zero error; MAPE can exceed 100% and is not an accuracy score. Zero actual values make the metric undefined, so this tool rejects them rather than silently dropping rows. Actual values near zero can dominate the result. Signed actual values are accepted using an absolute denominator, but MAPE is usually most interpretable for positive quantities with a meaningful zero. Use consistent units and compare models on the same observations. Enter up to 10000 pairs, one per line, using commas, semicolons, or spaces; omit thousands separators. Nonzero magnitudes must be between 1e-100 and 1e100. Scientific notation is supported. Compensated summation reduces rounding error; results show up to 12 significant digits. All processing stays in your browser.
Frequently Asked Questions
Everything you need to know about this tool, how it works, and privacy.
How is MAPE calculated?
Average the absolute errors divided by the absolute actual values, then multiply by 100. For actual 100, 200, 50 and predicted 90, 220, 50, MAPE is approximately 6.66666666667%.
Why are zero actual values rejected?
MAPE divides each error by its actual value. Division by zero is undefined, even when the prediction is also zero. This calculator rejects the input rather than omitting observations.
Can MAPE exceed 100%?
Yes. Predictions can differ from actual values by more than the actual magnitude. MAPE is an error measure, and subtracting it from 100 does not produce a general accuracy score.
When can MAPE be misleading?
Small actual values amplify percentage errors. MAPE is usually most useful for positive quantities with a meaningful zero. Signed values are supported mathematically but may be hard to interpret. Consider MAE or RMSE when zero or near-zero actual values occur.
Are my data uploaded?
No. Up to 10000 pairs are processed entirely in your browser. Use consistent units for actual and predicted values.