proc hpsplit: Rolling window training and validation instead of k-fold validation

[Baolinhdo]

Hi,

I am building a decision tree model using proc hpsplit. As I am dealing with time-series data, I want to do a walk-forward validation as suggested instead of 10-fold cross-validation or random sampling as validation set. Validation of the trained decision tree model is done in sliding window:

1. Starting at the beginning of the time series, 10-year window is used to train a model.
2. The model makes a prediction for the next year.
3. The prediction is stored or evaluated against known value
4. Next sliding 10-year window to include the known value and the process is repeated (back to step 1.)

How can we do this in SAS?

Thank you.

[ChrisNZ · Accepted solution]

Re-titled and moved to procedures thread.

[Baolinhdo]

Thanks, I will