I have a classification model that does not use the default threshold of 0.5, it uses a threshold of 0.3 and so I use the results from predict_proba() instead of
predict_proba()