/
azure.ai.ml.automl.ForecastingSettings.yml
751 lines (471 loc) · 17.4 KB
/
azure.ai.ml.automl.ForecastingSettings.yml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
### YamlMime:PythonClass
uid: azure.ai.ml.automl.ForecastingSettings
name: ForecastingSettings
fullName: azure.ai.ml.automl.ForecastingSettings
module: azure.ai.ml.automl
inheritances:
- azure.ai.ml.entities._mixins.RestTranslatableMixin
summary: Forecasting settings for an AutoML Job.
constructor:
syntax: 'ForecastingSettings(*, country_or_region_for_holidays: str | None = None,
cv_step_size: int | None = None, forecast_horizon: str | int | None = None, target_lags:
str | int | List[int] | None = None, target_rolling_window_size: str | int | None
= None, frequency: str | None = None, feature_lags: str | None = None, seasonality:
str | int | None = None, use_stl: str | None = None, short_series_handling_config:
str | None = None, target_aggregate_function: str | None = None, time_column_name:
str | None = None, time_series_id_column_names: str | List[str] | None = None,
features_unknown_at_forecast_time: str | List[str] | None = None)'
parameters:
- name: country_or_region_for_holidays
description: 'The country/region used to generate holiday features. These should
be ISO
3166 two-letter country/region code, for example ''US'' or ''GB''.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:str>]
- name: cv_step_size
description: 'Number of periods between the origin_time of one CV fold and the
next fold. For
example, if *n_step* = 3 for daily data, the origin time for each fold will
be
three days apart.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:int>]
- name: forecast_horizon
description: 'The desired maximum forecast horizon in units of time-series frequency.
The default value is 1.
Units are based on the time interval of your training data, e.g., monthly, weekly
that the forecaster
should predict out. When task type is forecasting, this parameter is required.
For more information on
setting forecasting parameters, see [Auto-train a time-series forecast model](https://docs.microsoft.com/azure/machine-learning/how-to-auto-train-forecast).'
isRequired: true
types:
- <xref:typing.Optional>[<xref:typing.Union>[<xref:int>, <xref:str>]]
- name: target_lags
description: "The number of past periods to lag from the target column. By default\
\ the lags are turned off.\n\nWhen forecasting, this parameter represents the\
\ number of rows to lag the target values based\non the frequency of the data.\
\ This is represented as a list or single integer. Lag should be used\nwhen\
\ the relationship between the independent variables and dependent variable\
\ do not match up or\ncorrelate by default. For example, when trying to forecast\
\ demand for a product, the demand in any\nmonth may depend on the price of\
\ specific commodities 3 months prior. In this example, you may want\nto lag\
\ the target (demand) negatively by 3 months so that the model is training on\
\ the correct\nrelationship. For more information, see [Auto-train a time-series\
\ forecast model](https://docs.microsoft.com/azure/machine-learning/how-to-auto-train-forecast).\n\
\n**Note on auto detection of target lags and rolling window size.\nPlease see\
\ the corresponding comments in the rolling window section.**\nWe use the next\
\ algorithm to detect the optimal target lag and rolling window size.\n\n1.\
\ Estimate the maximum lag order for the look back feature selection. In our\
\ case it is the number of periods till the next date frequency granularity\
\ i.e. if frequency is daily, it will be a week (7), if it is a week, it will\
\ be month (4). That values multiplied by two is the largest possible values\
\ of lags/rolling windows. In our examples, we will consider the maximum lag\
\ order of 14 and 8 respectively). \n\n2. Create a de-seasonalized series by\
\ adding trend and residual components. This will be used in the next step.\
\ \n\n3. Estimate the PACF - Partial Auto Correlation Function on the on the\
\ data from (2) and search for points, where the auto correlation is significant\
\ i.e. its absolute value is more then 1.96/square_root(maximal lag value),\
\ which correspond to significance of 95%. \n\n4. If all points are significant,\
\ we consider it being strong seasonality and do not create look back features.\
\ \n\n5. We scan the PACF values from the beginning and the value before the\
\ first insignificant auto correlation will designate the lag. If first significant\
\ element (value correlate with itself) is followed by insignificant, the lag\
\ will be 0 and we will not use look back features."
isRequired: true
types:
- <xref:typing.Union>[<xref:str>, <xref:int>, <xref:typing.List>[<xref:int>]]
- name: target_rolling_window_size
description: 'The number of past periods used to create a rolling window average
of the target column.
When forecasting, this parameter represents *n* historical periods to use to
generate forecasted values,
<= training set size. If omitted, *n* is the full training set size. Specify
this parameter
when you only want to consider a certain amount of history when training the
model.
If set to ''auto'', rolling window will be estimated as the last
value where the PACF is more then the significance threshold. Please see target_lags
section for details.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:typing.Union>[<xref:str>, <xref:int>]]
- name: frequency
description: 'Forecast frequency.
When forecasting, this parameter represents the period with which the forecast
is desired,
for example daily, weekly, yearly, etc. The forecast frequency is dataset frequency
by default.
You can optionally set it to greater (but not lesser) than dataset frequency.
We''ll aggregate the data and generate the results at forecast frequency. For
example,
for daily data, you can set the frequency to be daily, weekly or monthly, but
not hourly.
The frequency needs to be a pandas offset alias.
Please refer to pandas documentation for more information:
[https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#dateoffset-objects](https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#dateoffset-objects)'
isRequired: true
types:
- <xref:typing.Optional>[<xref:str>]
- name: feature_lags
description: Flag for generating lags for the numeric features with 'auto' or
None.
isRequired: true
types:
- <xref:typing.Optional>[<xref:str>]
- name: seasonality
description: 'Set time series seasonality as an integer multiple of the series
frequency.
If seasonality is set to ''auto'', it will be inferred.
If set to None, the time series is assumed non-seasonal which is equivalent
to seasonality=1.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:typing.Union>[<xref:int>, <xref:str>]]
- name: use_stl
description: 'Configure STL Decomposition of the time-series target column.
use_stl can take three values: None (default) - no stl decomposition, ''season''
- only generate
season component and season_trend - generate both season and trend components.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:str>]
- name: short_series_handling_config
description: 'The parameter defining how if AutoML should handle short time series.
Possible values: ''auto'' (default), ''pad'', ''drop'' and None.
* **auto** short series will be padded if there are no long series,
otherwise short series will be dropped.
* **pad** all the short series will be padded.
* **drop** all the short series will be dropped".
* **None** the short series will not be modified.
If set to ''pad'', the table will be padded with the zeroes and
empty values for the regressors and random values for target with the mean
equal to target value median for given time series id. If median is more or
equal
to zero, the minimal padded value will be clipped by zero.
Input:
:::row:::
:::column:::
**Date**
:::column-end:::
:::column:::
**numeric_value**
:::column-end:::
:::column:::
**string**
:::column-end:::
:::column:::
**target**
:::column-end:::
:::row-end:::
:::row:::
:::column:::
2020-01-01
:::column-end:::
:::column:::
23
:::column-end:::
:::column:::
green
:::column-end:::
:::column:::
55
:::column-end:::
:::row-end:::
Output assuming minimal number of values is four:
:::row:::
:::column:::
**Date**
:::column-end:::
:::column:::
**numeric_value**
:::column-end:::
:::column:::
**string**
:::column-end:::
:::column:::
**target**
:::column-end:::
:::row-end:::
:::row:::
:::column:::
2019-12-29
:::column-end:::
:::column:::
0
:::column-end:::
:::column:::
NA
:::column-end:::
:::column:::
55.1
:::column-end:::
:::row-end:::
:::row:::
:::column:::
2019-12-30
:::column-end:::
:::column:::
0
:::column-end:::
:::column:::
NA
:::column-end:::
:::column:::
55.6
:::column-end:::
:::row-end:::
:::row:::
:::column:::
2019-12-31
:::column-end:::
:::column:::
0
:::column-end:::
:::column:::
NA
:::column-end:::
:::column:::
54.5
:::column-end:::
:::row-end:::
:::row:::
:::column:::
2020-01-01
:::column-end:::
:::column:::
23
:::column-end:::
:::column:::
green
:::column-end:::
:::column:::
55
:::column-end:::
:::row-end:::
**Note:** We have two parameters short_series_handling_configuration and
legacy short_series_handling. When both parameters are set we are
synchronize them as shown in the table below (short_series_handling_configuration
and
short_series_handling for brevity are marked as handling_configuration and handling
respectively).
:::row:::
:::column:::
**handling**
:::column-end:::
:::column:::
**handling configuration**
:::column-end:::
:::column:::
**resulting handling**
:::column-end:::
:::column:::
**resulting handlingconfiguration**
:::column-end:::
:::row-end:::
:::row:::
:::column:::
True
:::column-end:::
:::column:::
auto
:::column-end:::
:::column:::
True
:::column-end:::
:::column:::
auto
:::column-end:::
:::row-end:::
:::row:::
:::column:::
True
:::column-end:::
:::column:::
pad
:::column-end:::
:::column:::
True
:::column-end:::
:::column:::
auto
:::column-end:::
:::row-end:::
:::row:::
:::column:::
True
:::column-end:::
:::column:::
drop
:::column-end:::
:::column:::
True
:::column-end:::
:::column:::
auto
:::column-end:::
:::row-end:::
:::row:::
:::column:::
True
:::column-end:::
:::column:::
None
:::column-end:::
:::column:::
False
:::column-end:::
:::column:::
None
:::column-end:::
:::row-end:::
:::row:::
:::column:::
False
:::column-end:::
:::column:::
auto
:::column-end:::
:::column:::
False
:::column-end:::
:::column:::
None
:::column-end:::
:::row-end:::
:::row:::
:::column:::
False
:::column-end:::
:::column:::
pad
:::column-end:::
:::column:::
False
:::column-end:::
:::column:::
None
:::column-end:::
:::row-end:::
:::row:::
:::column:::
False
:::column-end:::
:::column:::
drop
:::column-end:::
:::column:::
False
:::column-end:::
:::column:::
None
:::column-end:::
:::row-end:::
:::row:::
:::column:::
False
:::column-end:::
:::column:::
None
:::column-end:::
:::column:::
False
:::column-end:::
:::column:::
None
:::column-end:::
:::row-end:::'
isRequired: true
types:
- <xref:typing.Optional>[<xref:str>]
- name: target_aggregate_function
description: "The function to be used to aggregate the time series target\n \
\ column to conform to a user specified frequency. If the\n target_aggregation_function\
\ is set, but the freq parameter\n is not set, the error is raised. The possible\
\ target\n aggregation functions are: \"sum\", \"max\", \"min\" and \"mean\"\
.\n\n* The target column values are aggregated based on the specified operation.\
\ Typically, sum is appropriate for most scenarios. \n\n* Numerical predictor\
\ columns in your data are aggregated by sum, mean, minimum value, and maximum\
\ value. As a result, automated ML generates new columns suffixed with the aggregation\
\ function name and applies the selected aggregate operation. \n\n* For categorical\
\ predictor columns, the data is aggregated by mode, the most prominent category\
\ in the window. \n\n* Date predictor columns are aggregated by minimum value,\
\ maximum value and mode. \n\n:::row:::\n:::column:::\n**freq**\n:::column-end:::\n\
:::column:::\n**target_aggregation_function**\n:::column-end:::\n:::column:::\n\
**Data regularityfixing mechanism**\n:::column-end:::\n:::row-end:::\n:::row:::\n\
:::column:::\nNone (Default)\n:::column-end:::\n:::column:::\nNone (Default)\n\
:::column-end:::\n:::column:::\nThe aggregation is notapplied. If the validfrequency\
\ can not bedetermined the error willbe raised.\n:::column-end:::\n:::row-end:::\n\
:::row:::\n:::column:::\nSome Value\n:::column-end:::\n:::column:::\nNone (Default)\n\
:::column-end:::\n:::column:::\nThe aggregation is notapplied. If the numberof\
\ data points compliantto given frequency gridis less then 90% these pointswill\
\ be removed, otherwisethe error will be raised.\n:::column-end:::\n:::row-end:::\n\
:::row:::\n:::column:::\nNone (Default)\n:::column-end:::\n:::column:::\nAggregation\
\ function\n:::column-end:::\n:::column:::\nThe error about missingfrequency\
\ parameteris raised.\n:::column-end:::\n:::row-end:::\n:::row:::\n:::column:::\n\
Some Value\n:::column-end:::\n:::column:::\nAggregation function\n:::column-end:::\n\
:::column:::\nAggregate to frequency usingprovided aggregation function.\n:::column-end:::\n\
:::row-end:::"
isRequired: true
types:
- <xref:str>
- name: time_column_name
description: 'The name of the time column. This parameter is required when forecasting
to specify the datetime
column in the input data used for building the time series and inferring its
frequency.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:str>]
- name: time_series_id_column_names
description: 'The names of columns used to group a timeseries.
It can be used to create multiple series. If time series id column names is
not defined or
the identifier columns specified do not identify all the series in the dataset,
the time series identifiers
will be automatically created for your dataset.'
isRequired: true
types:
- <xref:typing.Union>[<xref:str>, <xref:typing.List>[<xref:str>]]
- name: features_unknown_at_forecast_time
description: 'The feature columns that are available for training but unknown
at the time of forecast/inference.
If features_unknown_at_forecast_time is set to an empty list, it is assumed
that
all the feature columns in the dataset are known at inference time. If this
parameter is not set
the support for future features is not enabled.'
isRequired: true
types:
- <xref:typing.Optional>[<xref:typing.Union>[<xref:str>, <xref:typing.List>[<xref:str>]]]
keywordOnlyParameters:
- name: country_or_region_for_holidays
isRequired: true
- name: cv_step_size
isRequired: true
- name: forecast_horizon
isRequired: true
- name: target_lags
isRequired: true
- name: target_rolling_window_size
isRequired: true
- name: frequency
isRequired: true
- name: feature_lags
isRequired: true
- name: seasonality
isRequired: true
- name: use_stl
isRequired: true
- name: short_series_handling_config
isRequired: true
- name: target_aggregate_function
isRequired: true
- name: time_column_name
isRequired: true
- name: time_series_id_column_names
isRequired: true
- name: features_unknown_at_forecast_time
isRequired: true