Description
When performing undersampling, certain data observations are removed. It may become impossible to have every categorical level in all categorical variables represented in the training data.
Solution
Consider using an oversampling method, or use categorical variables with fewer unique classes.