forecastflowml.ForecastFlowML#

class forecastflowml.ForecastFlowML(id_col, group_col, date_col, target_col, date_frequency, max_forecast_horizon, model_horizon, model, categorical_cols=None, use_lag_range=0)[source]#

Create forecaster Instance

Parameters:

  • id_col (str) – Time series identifer column.

  • group_col (str) – Column to partition the dataframe.

  • date_col (str) – Date column.

  • target_col (str) – Target column.

  • date_frequency (str) – Date frequency of the dataframe.

  • model_horizon (int) – Forecast horizon for a single model.

  • max_forecast_horizon (int) – Maximum horizon to generate the forecast. Needs to be multiple of model_horizon.

  • model (sklearn.base.BaseEstimator) – Regressor compatible with scikit-learn API.

  • categorical_cols (Optional[List[str]]) – List of columns to treat as categorical.

  • use_lag_range (int) – Extra lag range to use in addition to allowed lag values.

Methods

cross_validate(df[, n_cv_splits, ...])

Time series cross validation predictions

get_feature_importance([df_model])

The feature importances.

grid_search(df, param_grid[, n_cv_splits, ...])

Grid search with time series cross validation.

predict(df[, trained_models, spark])

Make predictions

train(df[, local_result])

Train models

Attributes

model_

Trained models in pickled format