The accuracy of intermittent demand estimates

By A. Syntetos, John E. Boylan. 2005

527 citations recorded by the source corpus.

Intermittent demand appears sporadically, with some time periods showing no demand at all. In this paper, four forecasting methods, Simple Moving Average (SMA, 13 periods), Single Exponential Smoothing (SES), Croston's method, and a new method (based on Croston's approach) recently developed by the authors, are compared on 3000 real intermittent demand data series from the automotive industry. The mean signed and relative geometric root-mean-square errors are shown to meet the theoretical and practical requirements of intermittent demand, as do the Percentage Better and Percentage Best summary statistics based on these measures. These measures are subsequently applied in a simulation experiment. The out-of-sample comparison results indicate superior performance of the new method. In addition, the results show that the mean signed error is not strongly scale dependent and the relative geometric root-mean-square error is a well-behaved accuracy measure for intermittent demand.

Subfields: Management Science and Operations Research, Marketing, Renewable Energy, Sustainability and the Environment

View Paper: https://www.semanticscholar.org/paper/5b1a4bfa508b2dc23baef3c996b6747c7a38c214

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 Marketing

In Renewable Energy, Sustainability and the Environment

Citations in the corpus

Another Look at Forecast Accuracy Metrics for Intermittent Demand (2006, methodology): “For intermittent-demand data, Syntetos and Boylan (2005) recommend the use of GMAE, although they call it the GRMSE.”
Performance Metrics (Error Measures) in Machine Learning Regression, Forecasting and Prognostics: Properties and Typology (2018, methodology): “Note that RelRMSE is also known as Theil’s U or U2 (De Gooijer and Hyndman, 2006).”
Demand Forecasting in the Fashion Industry: A Review (2013, unknown): “A lumpy demand is variable, sporadic and nervous [29].”
EXPONENTIAL SMOOTHING: THE STATE OF THE ART, PART II (2006, unknown): “Syntetos et al. (2005) extended Johnston and Boylan’s work by developing rules based on the variability of order size and inter-arrival times for selecting from three methods: N-N, original Croston, and modified Croston in Eqs.”
Intermittent demand forecasts with neural networks (2013, methodology): “The inversion bias discussed by Syntetos and Boylan (2001) exists again and the respective modifications for the SyntetosBoylan approximation is:”
'Horses for Courses' in demand forecasting (2014, methodology): “For the methods designed specifically for intermittent demand (Croston, SBA, TSB), small values for the smoothing parameters were applied, as suggested by the literature (Syntetos and Boylan, 2005).”
25 Years of IIF Time Series Forecasting: A Selective Review (2005, methodology): “Syntetos & Boylan (2005) proposed using the Relative Mean Absolute Error (see Section 10), while Willemain et al. (2004) recommended using the probability-integral transform method of Diebold et al.”
Demand categorisation in a European spare parts logistics network (2009, background): “showed that Croston's estimator is biased and they proposed ( Syntetos and Boylan, 2005”
Improving forecasting by estimating time series structural components across multiple frequencies (2014, methodology): “(2011) performed a more extensive empirical evaluation of the effect of the aggregation level in the performances of the random Walk and the Syntetos and Boylan approximation of Croston’s method (Syntetos and Boylan, 2005).”
On the demand distributions of spare parts (2012, unknown): “Research in the area of forecasting and stock control for intermittent demand items has developed rapidly in recent years (e.g. Syntetos and Boylan 2005, Chatfield and Hayya 2007, Quintana and Leung 2007, Altay et al. 2008, Gutierrez et al. 2008) with new results implemented into software products…”
Forecasting model selection through out-of-sample rolling horizon weighted errors (2011, unknown): “…Models: Simple smoothing; Holt (Holt et al., 1960); HoltWinter (additive or multiplicative) (Winters, 1960); Croston (Croston, 1972); Syntetos-Boylan (Syntetos & Boylan, 2005) • Moving Average Models: Simple moving average; Weighted simple moving average • ARIMA Models (Box & Jenkins, 1970):…”
Mean-based error measures for intermittent demand forecasting (2013, methodology): “a baseline measure. The most popular baseline is RW, in which case RelRMSE is Theil’s U2 statistic [22] and logRelMSE is Thompson’s LMR measure [23]. RelMAE was recommended for intermittent demand by [18] and called CumMAE by [1]. It is unlikely for these measures to become inﬁnite, because the deno”
Retail forecasting: Research and practice (2019, methodology): “Techniques designed specifically for intermittent demand include Croston’s method (Croston, 1972), the Syntetos and Boylan method (Syntetos and Boylan, 2001), Levén and Segerstedt method (Levén and Segerstedt, 2004), Syntetos–Boylan approximation (SBA) method (Syntetos and Boylan, 2005), and TSB met”
Forecasting intermittent inventory demands: simple parametric methods vs. bootstrapping ☆ (2015, methodology): “Although Croston claims that hismethod is unbiased, Syntetos and Boylan (2001) show that the opposite is true and present an improved method that corrects for bias (Syntetos & Boylan, 2005).”