Boosting multi-step autoregressive forecasts

Souhaib Ben Taieb, Rob Hyndman
; Proceedings of the 31st International Conference on Machine Learning, PMLR 32(1):109-117, 2014.

Abstract

Multi-step forecasts can be produced recursively by iterating a one-step model, or directly using a specific model for each horizon. Choosing between these two strategies is not an easy task since it involves a trade-off between bias and estimation variance over the forecast horizon. Using a nonlinear machine learning model makes the tradeoff even more difficult. To address this issue, we propose a new forecasting strategy which boosts traditional recursive linear forecasts with a direct strategy using a boosting autoregression procedure at each horizon. First, we investigate the performance of the proposed strategy in terms of bias and variance decomposition of the error using simulated time series. Then, we evaluate the proposed strategy on real-world time series from two forecasting competitions. Overall, we obtain excellent performance with respect to the standard forecasting strategies.

Cite this Paper


BibTeX
@InProceedings{pmlr-v32-taieb14, title = {Boosting multi-step autoregressive forecasts}, author = {Souhaib Ben Taieb and Rob Hyndman}, booktitle = {Proceedings of the 31st International Conference on Machine Learning}, pages = {109--117}, year = {2014}, editor = {Eric P. Xing and Tony Jebara}, volume = {32}, number = {1}, series = {Proceedings of Machine Learning Research}, address = {Bejing, China}, month = {22--24 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v32/taieb14.pdf}, url = {http://proceedings.mlr.press/v32/taieb14.html}, abstract = {Multi-step forecasts can be produced recursively by iterating a one-step model, or directly using a specific model for each horizon. Choosing between these two strategies is not an easy task since it involves a trade-off between bias and estimation variance over the forecast horizon. Using a nonlinear machine learning model makes the tradeoff even more difficult. To address this issue, we propose a new forecasting strategy which boosts traditional recursive linear forecasts with a direct strategy using a boosting autoregression procedure at each horizon. First, we investigate the performance of the proposed strategy in terms of bias and variance decomposition of the error using simulated time series. Then, we evaluate the proposed strategy on real-world time series from two forecasting competitions. Overall, we obtain excellent performance with respect to the standard forecasting strategies.} }
Endnote
%0 Conference Paper %T Boosting multi-step autoregressive forecasts %A Souhaib Ben Taieb %A Rob Hyndman %B Proceedings of the 31st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2014 %E Eric P. Xing %E Tony Jebara %F pmlr-v32-taieb14 %I PMLR %J Proceedings of Machine Learning Research %P 109--117 %U http://proceedings.mlr.press %V 32 %N 1 %W PMLR %X Multi-step forecasts can be produced recursively by iterating a one-step model, or directly using a specific model for each horizon. Choosing between these two strategies is not an easy task since it involves a trade-off between bias and estimation variance over the forecast horizon. Using a nonlinear machine learning model makes the tradeoff even more difficult. To address this issue, we propose a new forecasting strategy which boosts traditional recursive linear forecasts with a direct strategy using a boosting autoregression procedure at each horizon. First, we investigate the performance of the proposed strategy in terms of bias and variance decomposition of the error using simulated time series. Then, we evaluate the proposed strategy on real-world time series from two forecasting competitions. Overall, we obtain excellent performance with respect to the standard forecasting strategies.
RIS
TY - CPAPER TI - Boosting multi-step autoregressive forecasts AU - Souhaib Ben Taieb AU - Rob Hyndman BT - Proceedings of the 31st International Conference on Machine Learning PY - 2014/01/27 DA - 2014/01/27 ED - Eric P. Xing ED - Tony Jebara ID - pmlr-v32-taieb14 PB - PMLR SP - 109 DP - PMLR EP - 117 L1 - http://proceedings.mlr.press/v32/taieb14.pdf UR - http://proceedings.mlr.press/v32/taieb14.html AB - Multi-step forecasts can be produced recursively by iterating a one-step model, or directly using a specific model for each horizon. Choosing between these two strategies is not an easy task since it involves a trade-off between bias and estimation variance over the forecast horizon. Using a nonlinear machine learning model makes the tradeoff even more difficult. To address this issue, we propose a new forecasting strategy which boosts traditional recursive linear forecasts with a direct strategy using a boosting autoregression procedure at each horizon. First, we investigate the performance of the proposed strategy in terms of bias and variance decomposition of the error using simulated time series. Then, we evaluate the proposed strategy on real-world time series from two forecasting competitions. Overall, we obtain excellent performance with respect to the standard forecasting strategies. ER -
APA
Ben Taieb, S. & Hyndman, R.. (2014). Boosting multi-step autoregressive forecasts. Proceedings of the 31st International Conference on Machine Learning, in PMLR 32(1):109-117

Related Material