This function returns a table that contains seasonality information from a time series prediction model. The seasonality information contains the automatically detected seasons or periods. If you have specified your own values for the seasonality, these values are not necessarily used during forecasting.
Parameters
- regModel
- A value of type DM_RegressionModel
Return Value
The return value is of type TABLE
| Column | Data type | Purpose |
|---|---|---|
| submodel | VARCHAR(128) | name of the sub-model that is used to calculate the prediction. Each PMML document can contain multiple time series models for different columns of a given table. |
| period | REAL | repetition period of the season. The unit depends on the input time column. In case of a numeric type, no unit is assumed. In case of TIME, the unit is "second" in case of DATE, the unit is "days" and in case of TIMESTAMP, the unit is also "second" |
| weight | REAL | total weight of the period in the frequency spectrum of the time series; the value is between 0 and 1; it represents the area under the frequency peaks of the period and its harmonics divided by the total area under the frequency spectrum. |
| harmonicWeight | REAL | part of 'weight' that is contributed by the higher harmonics. The value is between 0 and 'weight'. |