Calculation Results
Results appear here after validation.
Model Comparison
| Model | MAE | Weighted MAE | MSE | RMSE | MAPE | Median AE | Bias | Maximum Error | R² | Normalised MAE | Improvement |
|---|
Detailed Observation Results
| # | Label | Model | Actual | Predicted | Weight | Signed Error | Absolute Error | Percentage Error | Squared Error | MAE Contribution |
|---|
Actual Versus Predicted
Absolute Error Bars
Error Distribution
Residual Plot
Actual Versus Predicted Scatter
Cumulative Absolute Error
Model MAE Comparison
Interpretation
Step-by-Step Calculation
Calculator Inputs
Enter paired target and prediction values. Every model must use matching rows.
Prediction Models
Column Mapping
Calculation Options
Formula Used
MAE = (1 ÷ n) × Σ |actual − predicted|
MAE averages absolute prediction differences. Its unit matches the target variable.
How to Use
- Enter actual values or upload a CSV file.
- Add one or more prediction models.
- Enter optional labels and observation weights.
- Select normalisation and zero-handling options.
- Calculate, inspect charts, then export results.
Example Data
| Observation | Actual | Predicted | Absolute Error |
|---|---|---|---|
| 1 | 3 | 2.5 | 0.5 |
| 2 | -0.5 | 0 | 0.5 |
| 3 | 2 | 2 | 0 |
| 4 | 7 | 8 | 1 |
The total absolute error equals two. Dividing by four gives 0.5.
Frequently Asked Questions
What does MAE measure?
MAE measures average absolute prediction error. Lower values usually indicate better predictions.
Is MAE always non-negative?
Yes. Absolute differences prevent positive and negative errors cancelling.
What units does MAE use?
MAE uses the target variable's original unit. This supports practical interpretation.
How does MAE differ from RMSE?
RMSE squares errors before averaging. Large errors therefore receive stronger penalties.
Can MAE compare different target scales?
Raw MAE cannot compare unrelated scales fairly. Use normalised MAE instead.
What is weighted MAE?
Weighted MAE gives selected observations greater influence. Weights must remain non-negative.
How are zero actual values handled?
MAE handles zeros normally. Percentage metrics require a selected zero rule.
Can several models be compared?
Yes. Add prediction models or map multiple CSV prediction columns.
Does MAE reveal prediction direction?
No. Review mean bias and signed residuals for direction.