29
Had to choose between AutoML and a custom model, I picked the easy way out
Last month I had a project at work where we needed a churn prediction model. I spent two days debating AutoML versus building a custom neural net from scratch. I went with AutoML because my boss wanted results in under a week. It got us an 82% accuracy score, which was fine, but the model is a total black box. Now the client keeps asking how it makes decisions and I have no real answer. Has anyone else taken the quick option and then regretted the lack of control later? I'm wondering if I should rebuild this thing before we go live.
0 comments
Log in to join the discussion
Log In0 Comments
No comments yet
Be the first to share your thoughts on this discussion.