The M3-Competition: results, conclusions and implications

By Spyros Makridakis, M. Hibon. 2000

1,729 citations recorded by the source corpus.

Subfields: Economics and Econometrics, Management Science and Operations Research, Evaluation & Benchmarks, Time-Series & Forecasting

How researchers see this paper

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

In Economics and Econometrics

In Management Science and Operations Research

In Evaluation & Benchmarks

In Time-Series & Forecasting

Citations in the corpus

Time-series forecasting with deep learning: a survey (2020, background): “]. Prior to the M4 competition of 2018 [61], the prevailing wisdom was that sophisticated methods do not produce more accurate forecasts, and simple models with ensembling had a tendency to do better [59, 62, 63]. Two key reasons have been identiﬁed to explain the underperformance of machine learnin”
N-BEATS: Neural basis expansion analysis for interpretable time series forecasting (2019, background): “sMAPE (symmetric Mean Absolute Percentage Error) and MASE (Mean Absolute Scaled Error) are standard scale-free metrics in the practice of forecasting [13, 20]: whereas sMAPE scales the error by the average between the forecast and ground truth, the MASE scales by the average error of the naïve predi”
Forecasting Across Time Series Databases using Long Short-Term Memory Networks on Groups of Similar Series (2017, background): “This was a conclusion of the influential M3 forecasting competition held in 1999 (Makridakis and Hibon, 2000).”
Optimal combination forecasts for hierarchical time series (2011, background): “better forecasts than ARIMA models ( Makridakis and Hibon, 2000 ), even when the underlying”
Forecasting with a hybrid method utilizing data smoothing, a variation of the Theta method and shrinkage of seasonal factors (2018, background): “We empirically evaluate the new proposition in the M3-Competition data with very promising results in terms of forecast accuracy.”
The Tourism Forecasting Competition (2011, methodology): “One method that performed extremely well in the M3 competition (Makridakis and Hibon, 2000) was the Theta method (Assimakopoulos and Nikolopoulos, 2000), which was further analysed and described by Hyndman and Billah (2003).”
Automatic Time Series Forecasting: The forecast Package for R (2008, methodology): “For more extensive testing of forecasting methods, the fma package contains the 90 data sets from Makridakis, Wheelwright, and Hyndman (1998), the expsmooth package contains 24 data sets from Hyndman, Koehler, Ord, and Snyder (2008b), and the Mcomp package contains the 1001 time series from the M-co”
Significance Tests Harm Progress in Forecasting (2007, neutral): “Readers should be clear about the conclusions from the KFHS analysis of the M3-Competition.”
Forecasting for big data: Does suboptimality matter? (2017, methodology): “In the original study (Makridakis and Hibon, 2000), 24 methods and commercial packages were compared with regards to their forecasting performance.”
Forecasting the novel coronavirus COVID-19 (2020, background): “This family has shown good forecast accuracy over several forecasting competitions [7] [8] [9] and is especially suitable for short series.”
Forecasting and operational research: a review (2008, methodology): “However, its empirical performance is generally worse than a damped trend smoothing benchmark (see Makridakis and Hibon, 2000; Gardner, 2006).”
Real-Time Density Forecasts From Bayesian Vector Autoregressions With Stochastic Volatility (2011, background): “In general, exponential smoothing also is known to be effective for trend estimation and forecasting (Makridakis and Hibon 2000; Chatfield et al. 2001).”
Optimizing Theta Model for Monthly Data (2013, methodology): “The M3 forecasting competition (Makridakis & Hibon, 2000) is regarded as the most successful competition to date, with more than 20 participants coming from both academia and software companies.”