LazyProphet: Time Series Forecasting with LightGBM

It’s all about the features

When we think about boosted trees in the time series space, it is usually in regards to the M5 competition where a significant portion of the top ten entries were using LightGBM. However, when looking at the performance of boosted trees in the univariate case, where there is not a wealth of exogenous features to utilize, their performance has been…rough.

Until now.

Section 1: The Intro

First, I should …

data science forecasting gradient-boosting lightgbm machine learning time time series time series forecasting