A review and comparison of strategies for multi-step ahead time series forecasting based on the NN5 forecasting competition

By Souhaib Ben Taieb, Gianluca Bontempi, Amir F. Atiya, Antti Sorjamaa. 2011

700 citations recorded by the source corpus.

Multi-step ahead forecasting is still an open challenge in time series forecasting. Several approaches that deal with this complex problem have been proposed in the literature but an extensive comparison on a large number of tasks is still missing. This paper aims to fill this gap by reviewing existing strategies for multi-step ahead forecasting and comparing them in theoretical and practical terms. To attain such an objective, we performed a large scale comparison of these different strategies using a large experimental benchmark (namely the 111 series from the NN5 forecasting competition). In addition, we considered the effects of deseasonalization, input variable selection, and forecast combination on these strategies and on multi-step ahead forecasting at large. The following three findings appear to be consistently supported by the experimental results: Multiple-Output strategies are the best performing approaches, deseasonalization leads to uniformly improved forecast accuracy, and input selection is more effective when performed in conjunction with deseasonalization.

Subfields: Management Science and Operations Research, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Management Science and Operations Research

In Signal Processing

Citations in the corpus

Forecasting Across Time Series Databases using Long Short-Term Memory Networks on Groups of Similar Series (2017, methodology): “Furthermore, Ben Taieb et al. (2011) discuss the benefits of applying a MIMO strategy over a single-output forecasting strategy (recursive strategy) in multi-step forecasting.”
Deep Learning for Time Series Forecasting: The Electric Load Case (2019, methodology): “es having different layers -recurrent or not- can be considered as well. 4.5 Multi-Step Prediction Schemes There are ﬁve different architecture-independent strategies for multi-step ahead forecasting [58]: 9 A PREPRINT Recursive strategy (Rec) a single model is trained to perform a one-step ahead fo”
Recurrent Neural Networks for Time Series Forecasting: Current Status and Future Directions (2019, methodology): “Ben Taieb et al. (2012) perform an extensive research involving five different techniques for multi-step-ahead forecasting.”
Neural network ensemble operators for time series forecasting (2014, methodology): “…of selecting a single NN that may be susceptible to poor initial values (or model setup), it is preferable to consider a combination of diﬀerent NN models (Hansen and Salamon, 1990; Zhang and Berardi, 2001; Versace et al., 2004; Barrow et al., 2010; Crone et al., 2011; Ben Taieb et al., 2012).”
Predictability of monthly temperature and precipitation using automatic time series forecasting methods (2018, background): “Relevant reviews on multi-step ahead forecasting methodologies can be found in Chevillon (2007) and Taieb et al. (2012).”
Machine Learning Strategies for Time Series Forecasting (2012, background): “A detailed review and comparison of strategies for multi-step time series forecasting based on the local learning algorithm is presented in [7].”
A Bias and Variance Analysis for Multistep-Ahead Time Series Forecasting (2016, positive): “For example, [33] provided a review and a comparison of several multistep strategies using the”
A Review of Deep Learning Models for Time Series Prediction (2019, neutral): “The timestamp h could be 1 [18], [19], or any positive integer, of which is named as multi-step-ahead prediction [20], [21].”