Hierarchical_Forecast

class hana_ml.algorithms.pal.tsa.hierarchical_forecast.Hierarchical_Forecast(method=None, weights=None)

Hierarchical forecast algorithm forecast across the hierarchy (that is, ensuring the forecasts sum appropriately across the levels).

Parameters:
method{'optimal_combination', 'bottom_up', 'top_down'}, optional

Method for reconciling forecasts across hierarchy.

Default to 'optimal_combination'.

weights{'ordinary_least_squares', 'minimum_trace', 'weighted_least_squares'}, optional

Only valid when parameter method is 'optimal_combination'.

Default to 'ordinary_least_squares'.

Examples

Input Dataframes:

>>> orig_df.collect().head(5)
    Series  TimeStamp   Original  Residual
0    Total       1992  48.748080  0.058252
1    Total       1993  49.480469  0.236069
2    Total       1994  49.932384 -0.044405
3    Total       1995  50.240702 -0.188002
4    Total       1996  50.608464 -0.128558
>>> pred_df.collect().head(5)
    Series  TimeStamp          VALUE
0    Total       1993      54.711279
1    Total       1994      54.207598
2    Total       1995      54.703918
3    Total       1996      55.200238
4    Total       1997      55.696558
>>> stru_df.collect().head(5)
   Index   Series    NUM
0      1    Total      2
1      2        A      3
2      3        B      2
3      4       AA      0
4      5       AB      0

Create a Hierarchical_Forecast instance:

>>> hr = Hierarchical_Forecast(method='optimal_combination',
                               weights='minimum_trace')

Perform fit_predict() on the given DataFrames:

>>> stats_tbl, result_tbl = hr.fit_predict(orig_data=orig_df,
                                           pred_data=pred_df,
                                           stru_data=stru_df)

Output:

>>> result_tbl.collect().head(5)
    Series   TimeStamp      VALUE
0    Total        1993  48.862705
1    Total        1994  54.255631
2    Total        1995  54.663688
3    Total        1996  55.192436
4    Total        1997  55.719965
Attributes:
result_DataFrame

Forecast result.

stats_DataFrame

Statistics.

Methods

fit_predict(orig_data, pred_data, stru_data)

Apply hierarchical forecast to the input DataFrames.

fit_predict(orig_data, pred_data, stru_data, orig_name=None, orig_key=None, orig_endog=None, orig_residual=None, pred_name=None, pred_key=None, pred_endog=None)

Apply hierarchical forecast to the input DataFrames.

Parameters:
orig_dataDataFrame

DataFrame of original data.

pred_dataDataFrame

DataFrame of predictive data.

stru_dataDataFrame

DataFrame of structure data.

orig_namestr, optional

Name of the time series name column.

orig_keystr, optional

Name of the time stamp column.

orig_endogstr, optional

Name of the raw data column.

orig_residualstr, optional

Name of the residual value column.

pred_namestr, optional

Name of time series name column.

pred_keystr, optional

Name of the time stamp column.

pred_endogstr, optional

Name of the predictive raw data column.

Inherited Methods from PALBase

Besides those methods mentioned above, the Hierarchical_Forecast class also inherits methods from PALBase class, please refer to PAL Base for more details.