Another look at measures of forecast accuracy

By Rob J Hyndman, Anne B. Koehler. 2006

5,557 citations recorded by the source corpus.

Subfields: Economics and Econometrics, Management Science and Operations Research, Statistics and Probability, Evaluation & Benchmarks

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 Statistics and Probability

In Evaluation & Benchmarks

Citations in the corpus

Forecasting Across Time Series Databases using Long Short-Term Memory Networks on Groups of Similar Series (2017, methodology): “close to zero in the original data and/or the forecasts. To address some of these issues, we use as a second evaluation metric in our experiments the Mean Absolute Scaled Error (MASE), as proposed in Hyndman and Koehler (2006). The MASE is a scale independent error measure, which also oers interpret”
Optimal combination forecasts for hierarchical time series (2011, methodology): “For each series, we select an optimal ARIMA model using the automatic algorithm of Hyndman & Koehler (2006).”
The Tourism Forecasting Competition (2011, methodology): “We also consider the MASE (mean absolute scaled error), which was proposed by Hyndman and Koehler (2006) in order to overcome some of the limitations of the MAPE.”
Which Artificial Intelligence Algorithm Better Predicts the Chinese Stock Market? (2018, methodology): “RMSE and MAPE have been widely used to evaluate the predictive accuracy [37], [38].”
Automatic Time Series Forecasting: The forecast Package for R (2008, methodology): “where k = 1 if c 6= 0 and 0 otherwise, and L is the maximized likelihood of the model fitted to the differenced data (1 − B)(1 − B)yt. The likelihood of the full model for yt is not actually defined and so the value of the AIC for different levels of differencing are not comparable. One solution to ”
Increasing Trust in AI Services through Supplier's Declarations of Conformity (2018, background): “…diﬀerent time-series, such as mean absolute percentage error (MAPE) and its symmetric version Armstrong (1978, p. 348), and (3) and scaled error rates that are preferred for comparisons of timeseries originating from diﬀerent platforms or of diﬀer-ent nature, explained in Hyndman & Koehler (2006).”
A note on the Mean Absolute Scaled Error (2015, background): “Hyndman and Koehler (2006) provide an exhaustive list of loss functions of realizations and forecasts, and for convenience a concise summary is presented in Table 1.”
Toward Automatic Time-Series Forecasting Using Neural Networks (2012, background): “Hyndman and Koehler [61] conducted a comparison study on different measures of forecast accuracy.”
Advances in forecasting with neural networks? Empirical evidence from the NN3 competition on time series prediction (2011, methodology): “…multiple error metrics, including the original symmetric mean absolute percent error (sMAPE), the mean absolute scaled error (MASE), as proposed by Hyndman and Koehler (2006), and two non-parametric tests proposed by Koning, Franses, Hibon, and Stekler (2005) in a follow-up analysis of the…”
Different Approaches to Forecast Interval Time Series: A Comparison in Finance (2011, background): “The Mean Absolute Scaled Error (MASE) proposed in Hyndman and Koehler (2006) is defined as the average of the scaled errors in absolute value, |qt |, for the considered period.”
Bagging exponential smoothing methods using STL decomposition and Box–Cox transformation (2016, methodology): “Furthermore, we use also the mean absolute scaled error (MASE) as proposed by Hyndman and Koehler (2006). It is defined as the mean absolute error on the test set, scaled by the mean absolute error of a benchmark method on the training set.”