The Kmeans clustering examples suggest normalizing the dataset by using the “scale” function before calling kmeans. After kmeans computes the cluster center, for using the computed model on a new data set for prediction, wouldn’t the model “centers” need to be de-normalized (reverse of ‘scale’ function) on the computed model.
Thanks,
Kumar