Enter Model Outputs
Use matching positions for actual, predicted, and optional weight values.
Example Data Table
This example shows how the calculator evaluates absolute error across five observations.
| # | Actual | Predicted | Weight | Absolute Error |
|---|---|---|---|---|
| 1 | 32 | 30 | 1.0 | 2 |
| 2 | 28 | 29 | 1.5 | 1 |
| 3 | 35 | 33 | 1.0 | 2 |
| 4 | 30 | 31 | 0.5 | 1 |
| 5 | 40 | 37 | 2.0 | 3 |
Unweighted MAE: (2 + 1 + 2 + 1 + 3) / 5 = 1.8
Weighted MAE: (2×1 + 1×1.5 + 2×1 + 1×0.5 + 3×2) / (1 + 1.5 + 1 + 0.5 + 2) = 2.0
Formula Used
Mean Absolute Error:
MAE = ( Σ | Predictedi − Actuali | ) / n
This metric measures the average absolute gap between predictions and true values.
Weighted Mean Absolute Error:
Weighted MAE = ( Σ wi × | Predictedi − Actuali | ) / Σ wi
Use weights when some observations deserve more importance than others.
Supporting Metrics:
MSE = Σ(Error²) / n
RMSE = √MSE
Mean Error = Σ(Error) / n
MAPE and sMAPE are shown when percent-based evaluation is meaningful.
How to Use This Calculator
- Enter a model name so your export files remain organized.
- Paste actual values into the first box.
- Paste predicted values into the second box using the same order.
- Optionally add weights for weighted MAE analysis.
- Choose your preferred decimal precision.
- Submit the form to display results above the calculator.
- Review summary metrics and the pair-by-pair error table.
- Download the CSV or PDF report for reporting, validation, or audit records.
Frequently Asked Questions
1. What does MAE measure in machine learning?
MAE measures the average absolute difference between actual values and predicted values. It shows how far predictions miss the target, without letting positive and negative errors cancel each other.
2. Why use MAE instead of RMSE?
MAE treats every error linearly, so it is easier to interpret in original units. RMSE gives heavier penalties to larger errors, which can be helpful when large misses are especially costly.
3. When should I use weighted MAE?
Use weighted MAE when some observations matter more than others. This is common in forecasting, pricing, healthcare, or operations where certain records carry higher business importance.
4. Can this calculator handle percentages too?
Yes. It also reports MAPE and sMAPE when calculations are valid. These help compare error in percentage terms, especially when actual values have different scales.
5. What happens if an actual value is zero?
MAE still works normally because it uses absolute differences. MAPE may become unavailable for those rows because percentage error divides by the actual value.
6. Does a lower MAE always mean a better model?
Usually yes, when comparing models on the same dataset and target scale. However, you should also review bias, RMSE, data context, and business constraints before choosing a final model.
7. What format should I use for input data?
You can separate numbers by lines, commas, spaces, semicolons, or vertical bars. Actual, predicted, and optional weight lists must align in the same sequence.
8. What do the CSV and PDF exports include?
The exports include the main summary metrics and the detailed row-level error table. This makes the file useful for validation, reporting, documentation, and stakeholder review.