Skip to main content

LightGBM Classifier

LightGBM Classifier is a tree-based ensemble model. It buckets continuous feature values into discrete bins and uses a gradient-based one-side sampling to split values. In comparison with XGBoost and CatBoost, LightGBM performs poorly on categorical datasets but has a similar performance to XGBoost on numerical datasets with fewer training time.

Advantages:​

  • It is computational efficient
  • It performs well on large sized datasets

Disadvantages:​

  • It has the problem of overfitting especially on small data