LightGBM, short for light gradient-boosting machine, is a free and open-source distributed gradient-boosting framework for machine learning initially developed by Microsoft. It uses decision tree algorithms for ranking, classification, and other machine-learning tasks.