In order to test the predictive accuracy of the models, you need to separate the outcome variable ("Div") in the testing data set from its predictors. You can do this with 1 line of code similar to the following: test.actual <- test.div[,1]