boosters.Metric#

class boosters.Metric#

Bases: object

Evaluation metrics for gradient boosting.

Each variant represents a different metric for evaluating model performance. Use the static constructor methods for validation.

Regression:

Metric.Rmse(): Root Mean Squared Error
Metric.Mae(): Mean Absolute Error
Metric.Mape(): Mean Absolute Percentage Error

Classification:

Metric.LogLoss(): Binary cross-entropy
Metric.Auc(): Area Under ROC Curve
Metric.Accuracy(): Classification accuracy

Note: ranking metrics are not implemented in core yet.

Examples

>>> from boosters import Metric
>>> metric = Metric.rmse()  # Regression
>>> metric = Metric.auc()  # Binary classification
>>> metric = Metric.Accuracy(threshold=0.7)

Pattern matching: >>> match metric: … case Metric.Rmse(): … print(“RMSE”) … case Metric.Accuracy(threshold=t): … print(f”Accuracy@{t}”)

Accuracy#: alias of PyMetric_Accuracy

Auc#: alias of PyMetric_Auc

LogLoss#: alias of PyMetric_LogLoss

Mae#: alias of PyMetric_Mae

Mape#: alias of PyMetric_Mape

MarginAccuracy#: alias of PyMetric_MarginAccuracy

MulticlassAccuracy#: alias of PyMetric_MulticlassAccuracy

MulticlassLogLoss#: alias of PyMetric_MulticlassLogLoss

None#: alias of PyMetric_None

PoissonDeviance#: alias of PyMetric_PoissonDeviance

Quantile#: alias of PyMetric_Quantile

Rmse#: alias of PyMetric_Rmse

static accuracy_at(threshold=0.5)#

Create accuracy metric with validation.

Use Metric.Accuracy() for default threshold, Metric.Accuracy(threshold=…) for direct construction, or this helper for validation.

static auc()#: Create AUC metric.

static logloss()#: Create log loss metric.

static mae()#: Create MAE metric.

static mape()#: Create MAPE metric.

static margin_accuracy()#: Create margin accuracy metric.

static multiclass_accuracy()#: Create multiclass accuracy metric.

static multiclass_logloss()#: Create multiclass log loss metric.

static none()#: Create no-metric.

static poisson_deviance()#: Create Poisson deviance metric.

static quantile(alpha)#: Create quantile metric with validation.

static rmse()#: Create RMSE metric.