Can not use model loaded into workspace

Question

Hi,

I load a trained model into the workspace using R, and when I use xgboost to predict, it always says:

         feature names stored in object and newdata are different

I checked the data for training and test, column names and order are the same, but test data has an extra column as a response/depending variable used for predicted results or compare prediction with it.

Without it, how do you know what to predict?

#modelclk$feature_names

test <- as.matrix(test2)

test$feature_names   ***(it says atomic vectors)***

dtest<- xgb.DMatrix(test2, missing=NaN)

pred<- predict(loadedmodel, dtest)

What is wrong?

A.

Answer 1

Ensure that the response/dependent variable column is not included in the matrix you pass to xgb.DMatrix. The matrix should only contain the feature columns.

Answer 2

Thanks, it works!

For old xxx.model, it always says error use model <- xxx.model. Is there anyway to load it?

Thanks,

N.