Top 5 things you should know about LightGBM

Important things you should pay attention to when using LightGBM

Top 5 key points covered in this article

We will talk about LightGBM on the following topics :

• Tree growing in LightGBM (Newton tree boosting, Histogram-based algorithm, node splitting and leaf-wise tree growth)
• Missing data handling
• Categorical feature handling
• Important hyperparameters
• Regularisations in LightGBM

Additional points covered in this article (if you are interested)

• Exclusive feature bundling
• Pairwise linear regression at each leaf
• Gradient-based one side sampling (GOSS)

🚀 Subscribe to us @ newsletter.verticalsolution.io for deep-dive content in DS, ML or AI! 🚀

Tree growing in LightGBM

In summary, LightGBM uses Newton tree boosting. During each tree growing, it supports faster histogram-based node splitting with leaf-wise growth.

Newton tree boosting