Search code examples
python-3.xlightgbm

LightGBM specify multiple metrics


What happens when I train a lightgbm model with multiple metrics?

enter image description here

I set 3 metrics and it turns out the best iteration result as above. But as you can see, even comparing with the last iteration, it does not seem to be the best result. I have check lightgbm documentation, and it only says the algo would minimise all metrics, but don't know how.

So how it works when minimising multiple metrics and why my result does not look right?


Solution

  • As it can be seen in the LightGBM documentation,

    early_stopping_round 🔗︎, default = 0, type = int, aliases: early_stopping_rounds, early_stopping

    will stop training if one metric of one validation data doesn’t improve in last early_stopping_round rounds
    

    And your AUC, which is a "higher better" metric, is lower at round 278 than it is at round 178. You should select the metric relevant to your problem to solve this issue : will you use your model for scoring or for classification?