The algorithms of APS 3.0 used for Time Series Prediction
allow not only the evolution of models but also to use these models to make predictions.
APS 3.0 allows you to make two kinds of predictions: one for testing past known events and another for making predictions about the future. In both cases, though, predictions are made recursively, by evaluating the forecast at
t+1, then using it to forecast t+2, and so on.
You must choose either one of these methods while loading your time series
data, as this imposes some constraints on the restructuring of the time series for training.
Namely,
n testing records (the n last ones) are saved for
testing and sometimes a small number of records from the top must be
deleted. However, you can also change these parameters later in the
APS modeling environment.
Thus, the first type of prediction can be used for research or pre-evaluation purposes, as it allows you to test the forecasting capabilities of your model on a set of test observations.
The second type is used obviously to predict unknown behavior, and APS 3.0 allows you to venture into the future as far as you see fit, by setting the number of predictions you want to make and then click the Predict button.
|