gluonts.dataset.stat module#

class gluonts.dataset.stat.DatasetStatistics(integer_dataset: bool, max_target: float, mean_abs_target: float, mean_target: float, mean_target_length: float, max_target_length: int, min_target: float, feat_static_real: List[Set[float]], feat_static_cat: List[Set[int]], num_past_feat_dynamic_real: Optional[int], num_feat_dynamic_real: Optional[int], num_feat_dynamic_cat: Optional[int], num_missing_values: int, num_time_observations: int, num_time_series: int, scale_histogram: gluonts.dataset.stat.ScaleHistogram)[source]#

Bases: tuple

A NamedTuple to store the statistics of a Dataset.

property feat_static_cat#: Alias for field number 8

property feat_static_real#: Alias for field number 7

property integer_dataset#: Alias for field number 0

property max_target#: Alias for field number 1

property max_target_length#: Alias for field number 5

property mean_abs_target#: Alias for field number 2

property mean_target#: Alias for field number 3

property mean_target_length#: Alias for field number 4

property min_target#: Alias for field number 6

property num_feat_dynamic_cat#: Alias for field number 11

property num_feat_dynamic_real#: Alias for field number 10

property num_missing_values#: Alias for field number 12

property num_past_feat_dynamic_real#: Alias for field number 9

property num_time_observations#: Alias for field number 13

property num_time_series#: Alias for field number 14

property scale_histogram#: Alias for field number 15

class gluonts.dataset.stat.ScaleHistogram(base: float = 2.0, bin_counts: Optional[dict] = None, empty_target_count: int = 0)[source]#

Bases: object

Scale histogram of a timeseries dataset.

This counts the number of timeseries whose mean of absolute values is in the [base ** i, base ** (i+1)] range for all possible i. The number of entries with empty target is counted separately.

Parameters

base – Log-width of the histogram’s buckets.
bin_counts –
empty_target_count –

add(target_values)[source]#

bucket_index(target_values)[source]#

count(target)[source]#

gluonts.dataset.stat.calculate_dataset_statistics(ts_dataset: Any) → gluonts.dataset.stat.DatasetStatistics[source]#

Computes the statistics of a given Dataset.

Parameters: ts_dataset – Dataset of which to compute the statistics.
Returns: NamedTuple containing the statistics.
Return type: DatasetStatistics