110625: All categorical classes of <value> unable to be sampled in Training Output.

Description

When randomly splitting data into training and testing sets, all categories must be present in the training set. The message displays any categorical variable that did not have all categories in the training set after 30 attempts. This error is most common when some categories have very few records and the size of the training set is small compared to the size of the testing set.

Solution

Identify any categories that have few records and either remove them or combine them with another category. Alternatively, increase the size of the training set by decreasing the Percent of Data as Test Subset parameter value.