A machine learning approach for forecasting hierarchical time series
作者：
Highlights：
• A novel strategy for forecasting hierarchical time series using is proposed.
• A deep neural network is used to extract time series features from the hierarchy.
• Explanatory variables are used to increase the forecasting accuracy of the hierarchy.
• Forecast reconciliation is performed at training time with a custom loss function.
• A new hierarchical sales dataset is presented.
摘要
•A novel strategy for forecasting hierarchical time series using is proposed.•A deep neural network is used to extract time series features from the hierarchy.•Explanatory variables are used to increase the forecasting accuracy of the hierarchy.•Forecast reconciliation is performed at training time with a custom loss function.•A new hierarchical sales dataset is presented.
论文关键词：Hierarchical time series,Forecast,Machine learning,Deep neural network
论文评审过程：Received 7 June 2020, Revised 14 April 2021, Accepted 21 April 2021, Available online 2 May 2021, Version of Record 29 May 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115102
