|
MaxHorizonFeaturizer
|
向 TimeSeriesDataSet 添加新行（高达最大预测边际）并还添加整数类型边际列的转换器。
示例：
raw_data = {'store': ['wholefoods'] * 4, ... 'date' : pd.to_datetime( ... ['2017-01-01', '2017-02-01', '2017-03-01', '2017-04-01']), ... 'sales': range(4)} tsds = TimeSeriesDataSet( ... data=pd.DataFrame(raw_data), ... time_series_id_column_names=['store'], time_column_name='date', ... target_colun_name='sales') tsds
sales
date store 2017-01-01 wholefoods 0 2017-02-01 wholefoods 1 2017-03-01 wholefoods 2 2017-04-01 wholefoods 3
>>> MaxHorizonFeaturizer(2).fit_transform(tsds).data
sales horizon_origin
date store origin
2017-01-01 wholefoods 2016-12-01 0 1
2016-11-01 0 2
2017-02-01 wholefoods 2017-01-01 1 1
2016-12-01 1 2
2017-03-01 wholefoods 2017-02-01 2 1
2017-01-01 2 2
2017-04-01 wholefoods 2017-03-01 3 1
2017-02-01 3 2
|