TD_ARIMAVALIDATE Function | Teradata Vantage - TD_ARIMAVALIDATE - Teradata Vantage

Database Unbounded Array Framework Time Series Functions

Deployment
VantageCloud
VantageCore
Edition
Enterprise
IntelliFlex
VMware
Product
Teradata Vantage
Release Number
17.20
Published
June 2022
Language
English (United States)
Last Update
2024-10-04
dita:mapPath
ncd1634149624743.ditamap
dita:ditavalPath
ruu1634160136230.ditaval
dita:id
ncd1634149624743
lifecycle
latest

The TD_ARIMAVALIDATE function performs in-sample forecasting for both seasonal and non-seasonal auto-regressive (AR), moving-average (MA), and ARIMA models. It also supports the extended Box-Jenkins seasonal ARIMA model formula.

ARIMA validation is the process of evaluating the accuracy and reliability of an ARIMA model for time series forecasting. The TD_ARIMAVALIDATE function provides metrics and residuals, such as mean absolute error (MAE), and mean squared error (MSE), that can be used for model selection and validation.

To perform ARIMA validation using TD_ARIMAVALIDATE, first train an ARIMA model on a training dataset using the appropriate parameters for the model, such as the order of differencing or autoregressive order. Once the model is trained, TD_ARIMAVALIDATE performs within-sample forecasting on the training dataset.

The amount of data processed by TD_ARIMAVALIDATE depends on the FIT_PERCENTAGE parameter used during the TD_ARIMAESTIMATE operation. For example, if FIT_PERCENTAGE was set to 80 during estimation, than that means that 80% of the series dataset was used for estimation processing, and 20% remains for the validation operation.

After evaluating the metrics, use the residuals produced by TD_ARIMAVALIDATE to perform additional model validation tests. Residuals are the differences between the predicted values and the actual values in the training dataset. Analyzing the residuals can identify any patterns or trends that the model may have missed, such as seasonality or trends, and determine if the model is correctly specified.

The output from TD_ARIMAVALIDATE includes the following optional layers:
  • ARTFITMETADATA layer containing metrics to assist with model selection
  • ARTFITMETADATA layer for goodness-of-fit
  • ARTFITRESIDUALS layer that can be used for performing statistical tests
The following procedure is an example of how to use TD_ARIMAVALIDATE:
  1. Run the TD_ARIMAESTIMATE function with a FIT_PERCENTAGE less than 100 to get the coefficients for the ARIMA model. The FIT_PERCENTAGE determines the portion of the input series to use for estimation and determines the portion (100 – FIT_PERCENTAGE) of the series to use for the validation operation.
  2. [Optional] Run TD_ARIMAVALIDATE function perform the validation operation on the remaining portion of the dataset.
  3. Run the TD_ARIMAFORECAST function with input from TD_ARIMAVALIDATE to forecast the future periods beyond the last observed period.