Gradient Boosted ARIMA for Time Series Forecasting

Boosting PmdArima’s Auto-Arima performance

TLDR: Adding gradient boosting to ARIMA adds complexity to the fitting procedure but can also drive accuracy if we optimize for new (p,d,q) parameters at each boosting round. Although, more gains can be achieved by boosting in conjunction with other methods.

All code lives here: ThymeBoost Github

For a full introduction to ThymeBoost view this article.

Introduction

Gradient Boosting has been a hot topic in the machine learning world for …

arima data science forecasting gradient python time time series time-series-analysis time series forecasting