Post by Bojan Tunguz

Starting with the version 1.6, XGBoost has been supporting multi-label regression and classification in the Python package. Internally the models are built one at a time, so the main advantage of this new functionality is in terms of API simplicity. However, there are also added benefits of reusing data and other integrated features like SHAP. Unfortunately, there have not been that many multi-label Kaggle competitions over the years. Nonetheless, I managed to dig up a couple for the sake of demonstration of this neat new XGBoost feature. Here is an example of multi-label classification with the original Toxic Comments competition and dataset: https://lnkd.in/gcpyUDSi And here is an example of multi-label regression from the July of 2021 Tabular Playground Series competition: https://lnkd.in/grEXYYR4 You can read the official documentation here: https://lnkd.in/gRGsC6Sj #DataScience #MachineLearning #ArtificialIntelligence #DS #ML #AI #XGBoost