Advances in forecasting with neural networks? Empirical evidence from the NN3 competition on time series prediction

By Sven F. Crone, M. Hibon, Konstantinos Nikolopoulos. 2011

278 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Management Science and Operations Research

How researchers see this paper

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

In Artificial Intelligence

In Management Science and Operations Research

Citations in the corpus

Forecasting Across Time Series Databases using Long Short-Term Memory Networks on Groups of Similar Series (2017, methodology): “ed modelling scheme to make it more desirable for large-scale business time series forecasting. There has been an increasing popularity of NNs with ensemble architectures for time series forecasting (Crone et al., 2011; Rahman et al., 2016; Barrow and Crone, 2016; Kourentzes et al., 2014b; Ben Taieb”
Layered Ensemble Architecture for Time Series Forecasting (2016, methodology): “We further evaluate the performance of layered bagging on the NN5 time series [38], another large benchmark time series TABLE VIII COMPARISON AMONG LAYERED BAGGING, YAN [28] AND TEN OTHER METHODS [60] BASED ON AVERAGE SMAPE, MASE, AND MDRAE.”
Recurrent Neural Networks for Time Series Forecasting: Current Status and Future Directions (2019, methodology): “In general, traditional univariate methods were on top in comparison to other computational intelligence methods at many forecasting competitions including NN3, NN5 and M3 (Crone et al., 2011; Makridakis and Hibon, 2000).”
Forecasting across time series databases using recurrent neural networks on groups of similar series: A clustering approach (2017, methodology): “In the NN3 competition (Crone et al., 2011), only one participating ML method was able to outperform damped trend exponential smoothing, and none of the methods was able to outperform the Theta method, which is equivalent to simple exponential smoothing with drift (Hyndman and Billah, 2003).”
Meta-learning with neural networks and landmarking for forecasting model selection an empirical evaluation of different feature sets applied to industry data (2016, methodology): “While statistical methods like exponential smoothing variants performed best in some studies [2], methods of computational intelligence (CI), such as neural networks, have shown promising performance in more recent studies [3].”
Toward Automatic Time-Series Forecasting Using Neural Networks (2012, methodology): “In this paper, the NN3 time-series competition dataset [12] is used for validating the ANN modeling scheme proposed.”
A comparison of AdaBoost algorithms for time series forecast combination (2016, methodology): “We assess all Boosting variants on the dataset of 111 empirical time series of industry sales from the NN3 forecasting competition (Crone, Hibon, and Nikolopoulos 2011).”
Forecasting: theory and practice (2020, background): “The NN3 competition (Crone et al., 2011) confirmed the superior performance of statistical methods, but noted that neural network approaches are closing the distance.”
'Horses for Courses' in demand forecasting (2014, methodology): “…facie variation of SES with drift, with the full theoretical underpinnings presented by Thomakos and 1 We use the term ‘factor’ to refer to both the data features as well as the strategic decision (forecasting horizon) that we will examine their impact on forecasting accuracy through this study.”
Optimizing Theta Model for Monthly Data (2013, background): “Towards this direction, many international forecasting competitions have conducted (for example Makridakis et al., 1982; Crone et al., 2011), aiming to provide an evaluation framework for well known and widely used approaches, such as exponential smoothing methods, versus new techniques and expert m”
Neural network ensemble operators for time series forecasting (2014, background): “Although the use of ensembles is nowadays accepted as the norm in forecasting with NNs (Crone et al., 2011), their performance is a function of how the individual forecasts are combined (Stock and Watson, 2004).”