<aside> 💡 LightGBM is considered to be a good baseline model
</aside>
The high-perfonnance LightGBM algorithm is capable of being distributed on multiple computers and handling large amounts of data quickly. It was developed by a team at Microsoft as an open-source project on GitHub.
LightGBM is based on Decision Trees, like XGBoost, but it follows a different strategy. While XGBoost uses decision trees to split on a variable and explore different tree splits at that variable (the level-wise tree growth strategy), LightGBM concentrates on one split and goes on splitting from there in order to achieve a better fit (the leaf-wise tree growth strategy). This allows LightGBM to quickly reach a good fit of the data, and to generate alternative solutions compared to XGBoost (which is good, if you expect to blend the two solutions together in order to reduce the variance of the estimates). Algorithmically speaking, if we think of the structure of splits operated by a decision tree as a graph, XGBoost pursues a breadth-first search (BPS) and LightGBM a depth-first search (DPS).