The Time Series Forecasting System is a point-and-click system that provides automatic model fitting and forecasting as well as interactive model development. The system provides a completely automatic forecasting model selection feature that selects the best-fitting model for each time series. Or, you can use system features to identify series behavior, fit candidate forecasting models, and perform diagnostic checks on the fitted models.
The Time Series Forecasting System includes a wide range of forecasting models. It provides tools to do the following:
fit forecasting models, select from a list of models, or build your own forecasting models and add them to the list of available models
use the automatic time series diagnostic facility to subset the available models list according to series properties of trend, seasonality, and need for log transformation, thus directing attention to the most promising models
decompose raw and transformed series variables and display the seasonally adjusted series, the trend-cycle component, the seasonal component, or the irregular component
Using graphical tools, you can display plots of predicted versus actual values.
You can explore plots and results with interactive graphical tools.
Plots
time series data
prediction errors
residual, partial, and inverse autocorrelations
significance probabilities
white noise and stationarity tests
forecasts with confidence limits
It is easy to view white noise tests and unit root tests.
Results
descriptive statistics
model parameters and significance estimates
model variance and final smoothing state
actual and forecasted values with prediction errors
log, logistic, square root, or Box-Cox transformations
changes in the series, including simple and seasonal differences and percent change
The model comparison window
You can choose between models and
compare the fit statistics for any two forecasting models side by side
customize the set of models fit by the automatic model selection process and customize the goodness-of-fit measure used for model selection
control the period of fit, period for evaluation of fit, and forecast horizon
use hold out samples to fit models to part of the data and evaluate predictions over the remainder of the data
print or save output files, including graphs, statistics of fit, parameters estimates, forecasts, and residuals
pave forecasting models and option settings for later reuse
review the history of the model selection process with an audit trail
print setup and print preview for graphics
perform automatic model fitting and forecasting from the command line
The intervention specification window