Metrics

Module with metrics of forecasting quality.

API details

Base:

Metric(metric_fn[, mode, metric_fn_signature])	Base class for all the multi-segment metrics.
MetricWithMissingHandling(metric_fn[, mode, ...])	Base class for all the multi-segment metrics that can handle missing values.

Enums:

MetricAggregationMode(value)	Enum for different metric aggregation modes.
MetricMissingMode(value)	Enum for different metric modes of working with missing values.

Scalar metrics:

MAE([mode, missing_mode])	Mean absolute error metric with multi-segment computation support.
MAPE([mode, missing_mode])	Mean absolute percentage error metric with multi-segment computation support.
MSE([mode, missing_mode])	Mean squared error metric with multi-segment computation support.
MSLE([mode, missing_mode])	Mean squared logarithmic error metric with multi-segment computation support.
R2([mode, missing_mode])	Coefficient of determination metric with multi-segment computation support.
RMSE([mode, missing_mode])	Root mean squared error metric with multi-segment computation support.
SMAPE([mode, missing_mode])	Symmetric mean absolute percentage error metric with multi-segment computation support.
WAPE([mode, missing_mode])	Weighted average percentage Error metric with multi-segment computation support.
MaxDeviation([mode, missing_mode])	Max Deviation metric with multi-segment computation support (maximum deviation value of cumulative sums).
MedAE([mode, missing_mode])	Median absolute error metric with multi-segment computation support.
Sign([mode, missing_mode])	Sign error metric with multi-segment computation support.
MissingCounter([mode])	Missing values counter with multi-segment computation support.

Interval metrics:

Coverage([quantiles, mode, upper_name, ...])	Coverage metric for prediction intervals - precenteage of samples in the interval [lower quantile, upper quantile].
Width([quantiles, mode, upper_name, ...])	Mean width of prediction intervals.

Utilities:

compute_metrics(metrics, y_true, y_pred)

Compute metrics for given y_true, y_pred.