Training a classification model on the Kaggle/StumbleUpon evergreen classification dataset