Metrics¶

Regression Metrics¶

Metric	Description	Range
`test_r2`	R-squared (coefficient of determination)	-∞ to 1 (higher is better)
`test_rmse`	Root mean squared error	0 to ∞ (lower is better)
`test_mae`	Mean absolute error	0 to ∞ (lower is better)
`test_mape`	Mean absolute percentage error	0% to ∞% (lower is better)

Classification Metrics¶

Metric	Description	Range
`test_accuracy`	Accuracy score	0 to 1 (higher is better)
`test_f1_macro`	F1 score (macro average)	0 to 1 (higher is better)
`test_f1_weighted`	F1 score (weighted average)	0 to 1 (higher is better)
`test_mcc`	Matthews correlation coefficient	-1 to 1 (higher is better)
`test_roc_auc`	ROC AUC score	0 to 1 (higher is better)

Accessing Metrics¶

ctx = run("data.csv", target="price")

# Access specific metric
print(ctx.metrics["test_r2"])

# Access all metrics
print(ctx.metrics)

# Model comparison
print(ctx.metrics["model_comparison"])

Overfitting Detection¶

open-mlpipe automatically detects overfitting:

Gap > 0.1: Warning — model may be overfitting
Gap < 0.01 and low score: Warning — model may be underfitting