LightGBM is an open-source gradient boosting library that builds decision trees quickly and with little memory, for classification, ranking and other tasks.
In plain words
Like XGBoost, LightGBM trains many decision trees one after another, each correcting the last. Its speed comes from shortcuts. It sorts each column's values into a small number of buckets instead of checking every exact value. It grows each tree at the leaf where splitting reduces the error most, rather than one full level at a time. Its original paper also skips many examples the model already handles well.
A quick example
Imagine a table with millions of rows of house sales. Instead of testing every exact value of a feature such as floor area, LightGBM groups the values into a limited number of ranges and only tests splits between those ranges. That makes each tree much cheaper to build.
Where you'll meet it
- Classification on large datasets.
- Ranking, such as ordering search results.
Related
Sources
- DocsLightGBM documentation, LightGBM developers · read 29 Sept 2026
- DocsFeatures, LightGBM developers · read 29 Sept 2026
- PaperLightGBM: A Highly Efficient Gradient Boosting Decision Tree, Ke et al. (NeurIPS 2017) · read 29 Sept 2026
- DocsLightGBM, LightGBM developers (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.