Accuracy measures: theoretical and practical concerns☆

By Spyros Makridakis. 1993

How researchers see this paper

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

Citations in the corpus

Forecasting stock market indexes using principle component analysis and stochastic time effective neural networks (2015, methodology): “There are different normalization methods that are tested to improve the network training [34,35], which include “the normalized data in the range of 1⁄20;1 ” in the following equation, which is also adopted in this work:”
EMD2FNN: A strategy combining empirical mode decomposition and factorization machine based neural network for stock market trend prediction (2019, background): “When the results are not consistent among the indexes, we choose the relatively more stable one, MAPE as suggested by Makridakis (Makridakis, 1993), to be the main reference substance.”
Simple robust averages of forecasts: Some empirical results (2008, methodology): “We base our evaluations on sMAPE rather than the simple MAPE (which entails dividing by Y instead of (X+Y) /2), because we feel that sMAPE is more stable and avoids some of the potential difficulties that can arise with MAPE (e.g., see Makridakis, 1993).”
A Modified CRITIC Method to Estimate the Objective Weights of Decision Criteria (2021, methodology): “Out of many metrics, sMAPE was chosen because unlike MAPE, it does not impose a larger penalty for negative error (when the estimated value is higher than actual value) than for positive error [61,62].”
Forecasting Time Series With Complex Seasonal Patterns Using Exponential Smoothing (2011, methodology): “Of the five accuracy measures reported by Makridakis et al. (1982), I use the mean absolute percentage error given by the following:”
HydroTest: A web-based toolbox of evaluation metrics for the standardised assessment of hydrological forecasts (2007, background): “(12)) is that the latter measurement is expressed in units that are relative to the observed record, as opposed to units that are relative to variation about the mean of the observed record, which could be difficult to interpret in an operational or decision-making context (Makridakis, 1993).”
Performance Metrics (Error Measures) in Machine Learning Regression, Forecasting and Prognostics: Properties and Typology (2018, background): “Initial intent behind this type of normalization was to make MAPE symmetric (Makridakis, 1993).”
The accuracy of intermittent demand estimates (2005, background): “The percentage of times better (PB) is easily interpreted (Makridakis, 1993) and is regarded as an intuitive nonparametric procedure (Makridakis & Hibon, 1995).”
MAPE-R: a rescaled measure of accuracy for cross-sectional subnational population forecasts (2011, unknown): “The symmetrical MAPE (SMAPE) was designed to deal with some of the limitations of the MAPE (Makridakis 1993).”
Error Metrics and Performance Fitness Indicators for Artificial Intelligence and Machine Learning in Engineering and Sciences (2020, unknown): “…s e l y a ff ec t e d i f a p r e d i c t e d v a l u e i s l a r g e r o r s m a ll e r t h a n t h e c o rr e s pond i ng ac t u a l v a l u e ) [ 49 ] • W o r k s w it h nu m e r i c d a t a 7 R e l a ti v e A b s o l u t e E rr o r ( R A E ) E xp r e ss e d a s a r a ti o c o m p a r i ng t h…”
25 Years of IIF Time Series Forecasting: A Selective Review (2005, background): “The “symmetric” measures were proposed by Makridakis (1993) in response to the observation that the MAPE and MdAPE have the disadvantage that they put a heavier penalty on positive errors than on negative errors.”