using the dataset ENRON-spam, 80/20 train/split, seed 42