The period for which data are held for testing a model. Let’s say we prepare a forecast of January 2017 by using the monthly data of January 2014 through December 2016. We have the data of January 2017, but we don’t include it because we want to use it to test the model. By comparing the actual of that period with the forecast, we can determine how good the model is. The January 2017 period in this case is the holdout period.