Specifying settings parameters for the Time Series mining function

For the different time series algorithms, general settings parameters can be specified.

Data warehousing in Db2® provides the following time series algorithms:

  • Autoregressive Integrated Moving Average (ARIMA)
  • Exponential Smoothing
  • Seasonal Trend Decomposition

In contrast to regression modeling, a time series run uses all algorithms by default and therefore incorporates all sub-models into a single PMML model. You can enable or disable the algorithms by using the functions DM_enableXXX or DM_disableXXX, where XXX can be replaced by ARIMA, ExpSm, or STD.

To specify settings for the time series mining function, you can specify the following parameters:

Time series time field
To set a value for the time or order column, use the DM_setTsTimeField parameter.

The time or order column must be of type TIME, DATE, or TIMESTAMP or of any other numerical type.

Time series target field
To add a field name of a time series target field of a DM_TsSettings object, use the DM_addTsTarget parameter

To remove an already set target field, use the DM_removeTsTarget parameter.

The mining data type of the target field must be numerical. The name of the target field must coincide with one of the field names that are defined in the logical data specifications.

Specify at least one time series target field for which a forecast should be computed.

From time
To specify the point in time from which the data is to be used for the analysis, use the DM_setFromTime parameter.
To time
To specify the point in time until the data is to be used for the analysis, use the DM_setToTime parameter.
Forecast horizon
To specify the point in time until the forecast is to be computed, use the DM_setForcHorizon parameter .
Seasonal cycle
To specify a manual seasonal cycle for each target field, use the DM_setSeasonalCyc parameter.
Interpolation method
To specify the way non-equidistant time series are interpolated, use the DM_setInterpolMeth parameter.
The following methods are provided:
  • Linear
  • Exponential splines
  • Cubic splines