Enter Model Metrics
Use three candidate models. Lower values are better for error, time, complexity, AIC, and BIC. Higher values are better for fit, validation, and stability.
Example Data Table
| Model | RMSE | R² | AIC | CV Score | Training Time | Complexity |
|---|---|---|---|---|---|---|
| Linear Baseline | 4.20 | 0.91 | 118.20 | 0.88 | 18 s | 6 |
| Regularized Model | 3.80 | 0.94 | 110.40 | 0.92 | 21 s | 8 |
| Compact Tree | 4.50 | 0.89 | 123.70 | 0.85 | 14 s | 5 |
Formula Used
- Min-max normalization: For higher-is-better metrics, normalized value = (x − min) ÷ (max − min).
- Cost normalization: For lower-is-better metrics, normalized value = (max − x) ÷ (max − min).
- Weighted composite score: Composite = [Σ(normalized metric × metric weight) ÷ Σ(weights)] × 100.
- Accuracy block: Mean of normalized R², adjusted R², and cross validation score, scaled to 100.
- Error control block: Mean of normalized RMSE, MAE, and MAPE, scaled to 100.
- Evidence block: Mean of normalized AIC, BIC, and stability, scaled to 100.
- Efficiency block: Mean of normalized training and prediction timing, scaled to 100.
How to Use This Calculator
- Enter names for three candidate models.
- Provide performance, fit, information, speed, complexity, and stability metrics for each model.
- Adjust weights to emphasize the measures that matter most in your study.
- Click Compare Models to generate the ranking table above the form.
- Review the best model, lead margin, and block scores for a balanced interpretation.
- Use the CSV or PDF buttons to export the ranked results.
Frequently Asked Questions
1. What does this tool compare?
This tool compares three mathematical or statistical models using weighted error, fit, efficiency, evidence, and stability metrics in one structured score.
2. Can I compare regression and classification models?
Yes, but only when you enter meaningful metrics for both. Keep your chosen metrics consistent so the ranking remains interpretable.
3. Why are some metrics treated as lower-is-better?
Error, time, complexity, AIC, and BIC usually improve when smaller. The tool inverts those measures during normalization before scoring.
4. What is the composite score?
The composite score is a weighted average of normalized metrics, scaled to 100. Higher values indicate stronger overall performance.
5. Why use weights?
Weights let you emphasize what matters most, such as prediction accuracy, simpler structure, lower latency, or stronger information criteria.
6. What happens if all models share one metric value?
That metric receives the same normalized score for all models, so it does not distort the final ranking.
7. Is a higher evidence block always better?
Yes within this framework, because it rewards lower AIC, lower BIC, and higher stability after normalization.
8. Should I select only the top score?
Usually yes, but also inspect the gap, efficiency, and complexity. Close scores may justify choosing a simpler or faster model.