Just add some details about my question. I have monthly time-series data. And I want to predict forward 3 months. So there are some overlapping periods when running cross-validation. For example, if Jan and Feb are in the training set and March in the testing set, the model might know what is likely to happen when using the training set due to the overlapped periods. Currently, I run sampling before building the tree model and assign group id for cross-validation. For example, assign Jan, Feb, and Mar as group 1 and Jun and Jul as group 2. Pass the group ID to the tree model function. I implement it in R. Not sure how SAS can do it.
... View more
