Netezza Time Series Build Options

There are two levels of build options:

  • Basic - settings for the algorithm choice, interpolation, and time range to be used.
  • Advanced - settings for forecasting

This section describes the basic options.

The Build Options tab is where you set all the options for building the model. You can, of course, just click the Run button to build a model with all the default options, but normally you will want to customize the build for your own purposes.

Algorithm

These are the settings relating to the time series algorithm to be used.

Algorithm Name. Choose the time series algorithm you want to use. The available algorithms are Spectral Analysis, Exponential Smoothing (default), ARIMA, or Seasonal Trend Decomposition. See the topic Netezza Time Series for more information.

Trend. (Exponential Smoothing only) Simple exponential smoothing does not perform well if the time series exhibits a trend. Use this field to specify the trend, if any, so that the algorithm can take account of it.

  • System Determined. (default) The system attempts to find the optimal value for this parameter.
  • None(N). The time series does not exhibit a trend.
  • Additive(A). A trend that steadily increases over time.
  • Damped Additive(DA). An additive trend that eventually disappears.
  • Multiplicative(M). A trend that increases over time, typically more rapidly than a steady additive trend.
  • Damped Multiplicative(DM). A multiplicative trend that eventually disappears.

Seasonality. (Exponential Smoothing only) Use this field to specify whether the time series exhibits any seasonal patterns in the data.

  • System Determined. (default) The system attempts to find the optimal value for this parameter.
  • None(N). The time series does not exhibit seasonal patterns.
  • Additive(A). The pattern of seasonal fluctuations exhibits a steady upward trend over time.
  • Multiplicative(M). Same as additive seasonality, but in addition the amplitude (the distance between the high and low points) of the seasonal fluctuations increases relative to the overall upward trend of the fluctuations.

Use system determined settings for ARIMA. (ARIMA only) Choose this option if you want the system to determine the settings for the ARIMA algorithm.

Specify. (ARIMA only) Choose this option and click the button to specify the ARIMA settings manually.

Interpolation

If the time series source data has missing values, choose a method for inserting estimated values to fill the gaps in the data. See the topic Interpolation of Values in Netezza Time Series for more information.

  • Linear. Choose this method if the intervals of the time series are regular but some values are simply not present.
  • Exponential Splines. Fits a smooth curve where the known data point values increase or decrease at a high rate.
  • Cubic Splines. Fits a smooth curve to the known data points to estimate the missing values.

Time Range

Here you can choose whether to use the full range of data in the time series, or a contiguous subset of that data, to create the model. Valid input for these fields is defined by the data storage type of the field specified for Time Points on the Fields tab. See the topic Netezza Time Series Field Options for more information.

  • Use earliest and latest times available in data. Choose this option if you want to use the full range of the time series data.
  • Specify time window. Choose this option if you want to use only a portion of the time series. Use the Earliest time (from) and Latest time (to) fields to specify the boundaries.