I’m doing an XGBRegressor and having issues with my final prediction because I’m somehow generating negative predictions when I expect a value from 0-1. My solution has been to just cut off my predictions at 0, but I think that might be hurting my overall score. Is there a better way I should be handling my predictions? Here is what I’m using currently:
test_predictions = m_xgb.predict(df_test_keep.values)
That gives me predictions that look like this:
So I’m just taking that and cutting off the bottom of it so it looks like this instead:
but I’m worried that it is giving me poor results. Any thoughts on a better way to handle this?