Explainable boosted linear regression for time series forecasting

作者：

Highlights：

• Explainable boosted linear regression (EBLR) is proposed for time series forecasting.

• EBLR starts with a base model, and explains model’s errors through regression trees.

• EBLR’s training residuals can be used to generate probabilistic forecasts.

• EBLR performs comparably well and is simpler/more interpretable than other methods.

摘要

•Explainable boosted linear regression (EBLR) is proposed for time series forecasting.•EBLR starts with a base model, and explains model’s errors through regression trees.•EBLR’s training residuals can be used to generate probabilistic forecasts.•EBLR performs comparably well and is simpler/more interpretable than other methods.

论文关键词：Time series regression,Probabilistic forecasting,Decision trees,Linear regression,ARIMA

论文评审过程：Received 18 September 2020, Revised 1 June 2021, Accepted 27 June 2021, Available online 8 July 2021, Version of Record 21 July 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108144