Pedro Tabacof - Unlocking the Power of Gradient-Boosted Trees (using LightGBM) | PyData London 2022

[npatta01]

Video Title: Pedro Tabacof - Unlocking the Power of Gradient-Boosted Trees (using LightGBM) | PyData London 2022

Video Link: https://www.youtube.com/watch?v=qGsHlvE8KZM&t=2s&ab_channel=PyData

Contents: 00:10 Introduction of speaker 00:15 Speaker begins talk 00:50 Why LightGBM 01:30 Speaker bio 02:20 Agenda 02:40 Example of LightGBM on Kaggle house priices prediction 04:43 Kaggle gradient-boosted trees performance 07:00 How gradient boosted trees work 09:36 What are the important tuning parameters 11:15 Handling missing values 13:41 Handling categorical features 15:35 Supported Loss functions 18:58 Custom loss functions 20:24 Early stopping 20:58 Sample / class weights 22:22 Other cool features in LightGBM 23:35 Interpretability: SHAP 25:06 Conclusion and takeway 27:59 Q&A