Regression Model Validation Metrics Calculator

Test models with error and fit scores. Review residual patterns through charts and summary tables. Download clean reports for validation reviews and stakeholder sharing.

Calculator Input

Model Name

Predictor Count

Decimal Places

Actual and Predicted Values

Use one row per prediction pair. A header row is allowed. Separate values with commas, spaces, tabs, semicolons, or pipes.

Example Data Table

Row	Actual	Predicted
1	120	118
2	132	130
3	128	131
4	145	141
5	150	149
6	160	158
7	170	172
8	175	173
9	180	182
10	190	188

This example shows a compact regression validation dataset. Load it instantly with the example button and calculate all metrics.

Formula Used

Residual: Predicted − Actual
MAE: Σ|Actual − Predicted| / n
MSE: Σ(Actual − Predicted)² / n
RMSE: √MSE
MAPE: [Σ|((Actual − Predicted) / Actual)| / n] × 100
sMAPE: Mean of |Actual − Predicted| / ((|Actual| + |Predicted|) / 2) × 100
WAPE: Σ|Actual − Predicted| / Σ|Actual| × 100
R²: 1 − SSE / SST
Adjusted R²: 1 − [(1 − R²)(n − 1) / (n − p − 1)]
Explained Variance: 1 − Var(Residuals) / Var(Actual)
Durbin-Watson: Σ(eᵢ − eᵢ₋₁)² / Σeᵢ²
RMSLE: √Mean[(ln(1 + Predicted) − ln(1 + Actual))²]
AIC: n ln(SSE / n) + 2k
BIC: n ln(SSE / n) + k ln(n)

How to Use This Calculator

Enter a model name for reporting clarity.
Set the number of predictors used by your regression model.
Paste actual and predicted values into the dataset box.
Keep one pair per line. Optional headers are allowed.
Choose your preferred decimal precision.
Click Calculate Metrics to generate validation results.
Review summary cards, the full metric table, and residual charts.
Export the report as CSV or PDF when needed.

FAQs

1. What does this calculator measure?

It measures regression model quality using common validation metrics. You can compare prediction error, goodness of fit, residual behavior, and complexity-adjusted performance from one dataset.

2. Why do I need actual and predicted values?

These two columns are the basis for nearly every validation metric. Their difference creates residuals, which then drive error, fit, and stability calculations.

3. What is a good RMSE value?

A good RMSE is small relative to the target scale. It should always be judged against business tolerance, target range, and competing models on the same dataset.

4. Why can MAPE show N/A?

MAPE divides by actual values. If actual values include zeros, percentage error becomes undefined for those rows. In such cases, MAE, RMSE, or sMAPE can be more reliable.

5. What does adjusted R² add?

Adjusted R² accounts for predictor count. It helps prevent overvaluing models that improve plain R² only by adding more features without enough real predictive gain.

6. When should I use RMSLE?

Use RMSLE when targets are nonnegative and you care more about relative growth differences than raw absolute misses. It is common for skewed targets.

7. What does Durbin-Watson tell me?

Durbin-Watson checks residual autocorrelation. Values near 2 suggest independence. Much lower or higher values can indicate serial patterns the model failed to capture.

8. Should I trust one metric alone?

No. Strong validation combines several metrics, residual charts, and domain context. A model can score well on one measure while failing stability or interpretability checks.

Row	Actual	Predicted
1	120	118
2	132	130
3	128	131
4	145	141
5	150	149
6	160	158
7	170	172
8	175	173
9	180	182
10	190	188

Row	Actual	Predicted
1	120	118
2	132	130
3	128	131
4	145	141
5	150	149
6	160	158
7	170	172
8	175	173
9	180	182
10	190	188