Contents

This Page

  1. Docs
  2. actableai.timeseries package

actableai.timeseries package

Subpackages

Submodules

actableai.timeseries.dataset module

class actableai.timeseries.dataset.AAITimeSeriesDataset(dataframes: Union[pandas.core.frame.DataFrame, pandas.core.series.Series, List[pandas.core.frame.DataFrame], Dict[str, pandas.core.frame.DataFrame]], target_columns: List[str], freq: Optional[str] = None, prediction_length: int = 0, training: bool = True, group_by: Optional[List[str]] = None, date_column: Optional[str] = None, feature_columns: Optional[Union[str, List[str]]] = None, feat_dynamic_real: Optional[List[str]] = None, feat_dynamic_cat: Optional[List[str]] = None, feat_static_real: Optional[List[str]] = None, feat_static_cat: Optional[List[str]] = None, seasonal_periods: Optional[List[int]] = None)

Bases: object

GluonTS-compatible custom class to store a time series dataset.

clean_features(keep_feat_static_real: bool, keep_feat_static_cat: bool, keep_feat_dynamic_real: bool, keep_feat_dynamic_cat: bool) → actableai.timeseries.dataset.AAITimeSeriesDataset

Filter out features from dataset.

Parameters
  • keep_feat_static_real – Whether to keep the real static features or not.
  • keep_feat_static_cat – Whether to keep the categorical static features or not.
  • keep_feat_dynamic_real – Whether to keep to the real dynamic features or not.
  • keep_feat_dynamic_cat – Whether to keep the categorical dynamic features or not.
Returns

Filtered Dataset.

property group_list: List[Tuple[Any, ...]]

Returns the list of group in the dataset.

Returns
List of group.
slice_data(slice_df: Optional[Union[slice, Callable]], copy: bool = False) → actableai.timeseries.dataset.AAITimeSeriesDataset

Slice dataset.

Parameters
  • slice_df – Slice or function to call that will return a slice. The slice will be applied to each group separately.
  • copy – If True the data will be copied.
Returns

The sliced dataset.

actableai.timeseries.forecast module

class actableai.timeseries.forecast.AAITimeSeriesForecast(forecast: gluonts.model.forecast.Forecast, transformation_func: Optional[Callable[[numpy.ndarray], numpy.ndarray]] = None)

Bases: gluonts.model.forecast.Forecast

Custom wrapper around GluonTS Forecast.

copy_aggregate(agg_fun: Callable) → actableai.timeseries.forecast.AAITimeSeriesForecast
Returns a new Forecast object with a time series aggregated over the
dimension axis.
Parameters
agg_fun – Aggregation function that defines the aggregation operation (typically mean or sum).
Returns
Selected sub-forecast.
copy_dim(dim: int) → actableai.timeseries.forecast.AAITimeSeriesForecast

Returns a new Forecast object with only the selected sub-dimension.

Parameters
dim – The returned forecast object will only represent this dimension.
Returns
Selected sub-forecast.
dim() → int

Returns the dimensionality of the forecast object.

Returns
Forecast’s dimension.
info: Optional[Dict]
item_id: Optional[str]
property mean: numpy.ndarray

Mean of the forecast.

Returns
Forecast’s mean.
prediction_length: int
quantile(q: Union[float, str]) → numpy.ndarray

Computes a quantile from the predicted distribution.

Parameters
q – Quantile to compute.
Returns
Value of the quantile across the prediction range.
start_date: pandas._libs.tslibs.period.Period
to_dataframe(target_columns: List[str], date_list: List[datetime.datetime], quantiles: List[float] = [0.05, 0.5, 0.95]) → pandas.core.frame.DataFrame

Convert GluonTS forecast to pandas DataFrame.

Parameters
  • target_columns – List of columns to forecast.
  • date_list – List of datetime forecasted.
  • quantiles – List of quantiles to forecast.
Returns

Forecasted values as pandas DataFrame.

actableai.timeseries.utils module

actableai.timeseries.utils.find_freq(pd_date: pandas.core.series.Series, period: int = 10) → Optional[str]

Find the frequency from a list of datetime. :param pd_date: List of datetime as a pandas Series, needs to be sorted. :param period: Window to look for when computing the frequency.

Returns
The frequency or None if not found.
actableai.timeseries.utils.find_gluonts_freq(pd_date: pandas.core.series.Series, freq: str) → str

Convert pandas frequency to GluonTS frequency.

Parameters
  • pd_date – List of datetime as a pandas Series.
  • freq – pandas frequency.
Returns

GluonTS frequency.

actableai.timeseries.utils.get_satisfied_formats(row: pandas.core.series.Series, unique_formats: List[str]) → List[str]

Find the datetime formats compatible with a list of datetime.

Parameters
  • row – List of datetime as a pandas Series.
  • unique_formats – List of formats to try.
Returns

List of compatible formats.

actableai.timeseries.utils.handle_datetime_column(series: pandas.core.series.Series, min_parsed_rate: float = 0.5) → Tuple[pandas.core.series.Series, str]

Parse datetime from a list of datetime.

Parameters
  • series – List of datetime to parse as a pandas Series.
  • min_parsed_rate – Ratio of minimum parsed dates.
Returns

  • Parsed pandas series.
  • Data type of the parsed series [“datetime”, “others”].

actableai.timeseries.utils.interpolate(df: pandas.core.frame.DataFrame, freq: str) → pandas.core.frame.DataFrame

Interpolate missing values in time series.

Parameters
  • df – Time series DataFrame.
  • freq – Frequency to use when interpolating
Returns

Interpolated new DataFrame

actableai.timeseries.utils.parse_by_format_with_valid_frequency(series: pandas.core.series.Series, formats: List[str]) → pandas.core.series.Series

Parse datetime using a list of formats. Returns the first working format.

Parameters
  • series – List of datetime as pandas Series.
  • formats – List of datetime formats to try.
Returns

Parsed list of datetime.

actableai.timeseries.utils.parse_datetime(dt_str: pandas.core.series.Series, formats: List[str]) → Optional[pandas.core.series.Series]

Try to parse datetime using a list of formats. Returns the first working format.

Parameters
  • dt_str – List of datetime as pandas Series.
  • formats – List of datetime formats to try.
Returns

Parsed list of datetime or None if no formats are compatible.

Module contents

© Copyright 2022, ActableAI Team. Created using Sphinx.