Time Series Table Support - Teradata Warehouse Miner

Teradata provides support for tables containing time series data. Time series tables typically contain a unique identifier for an object being measured over time, a date or timestamp that an observation is made, a sequence number to allow multiple observations for a given time, and one or more measurements taken at the time of an observation. A time series table is identified by a PRIMARY TIME INDEX clause in the CREATE TABLE statement that defines it. A time series table contains system generated columns of time information that are used in distributing data across the system. New time series aggregation functions are provided, along with a GROUP BY TIME clause that may be used with new, or in many cases, traditional aggregation functions.

The following is an outline of features in Teradata Warehouse Miner that provide time series support:

• PRIMARY TIME INDEX options on the OUTPUT—primary index tab
• GROUP BY TIME clause on the INPUT—expert options tab (SQL Elements category Other—Expert)
• Time series functions TD_GETTIMEBUCKET and TD_TIME_BUCKET_NUMBER (SQL Elements category XML, JSON etc.)
• Time series virtual columns $TD_TIMECODE_RANGE and $TD_GROUP_BY_TIME (SQL Elements category XML, JSON etc.)
• Time series macro TD_TIME_SERIES_RANGE (SQL Elements category Run Units)
• Aggregation functions (SQL ELEMENTS category Aggregation, sub-category Time Series)
  • Average
  • Bottom
  • Count
  • Delta_T
  • Describe
  • First
  • Kurtosis
  • Last
  • Maximum
  • Median
  • Median Absolute Deviation
  • Minimum
  • Mode
  • Percentile
  • Skewness
  • Standard Deviation
  • Sum
  • Top
  • Variance
• Even though TD_TIMEBUCKET is displayed as an available column on the INPUT—data selection tab, it should not be selected. It is one of the system generated time series columns and is used for the distribution of data on the system.
• In a similar way, the TD_TIMEBUCKET generated column should not be selected as an anchor column on the INPUT—anchor table tab. (Ordinarily, the anchor columns default to the primary index columns of the anchor table, but this is not the case when the anchor table is a time series table. In this case, the anchor columns are initially left unselected.)

For Variable Creation – INPUT – Expert Option details, refer to Teradata Warehouse Miner User Guide: Volume 2 ADS Generation.