Improving forecasting by estimating time series structural components across multiple frequencies

By Nikolaos Kourentzes, Fotios Petropoulos, Juan Ramón Trapero. 2014

188 citations recorded by the source corpus.

Subfields: Management Science and Operations Research, Signal Processing

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 Signal Processing

Citations in the corpus

Forecasting with a hybrid method utilizing data smoothing, a variation of the Theta method and shrinkage of seasonal factors (2018, background): “For instance, Kourentzes et al. (2014) argue that temporal aggregated series of low frequencies can be used for identifying long-term characteristics of the data, while Proietti and Lütkepohl (2013) suggest the use of power transformations for rescaling the historical data, therefore simplifying the”
'Horses for Courses' in demand forecasting (2014, methodology): “Nikolopoulos et al. (2011) and the ADIDA framework indicate that forecasters may via temporal aggregations switch to different frequencies (than the ones the data are observed), while Kourentzes et al. (2014) present a framework to efficiently combine forecasts derived from multiple frequencies.”
A simple combination of univariate models (2020, background): “• Multiple temporal aggregation (MTA): Many recent studies have shown the positive impact of multiple temporal aggregation on forecasting accuracy (for example, see: Kourentzes et al., 2014; Petropoulos and Kourentzes, 2014; Athanasopoulos et al., 2017).”
Forecasting: theory and practice (2020, background): “Instead of focusing on a single aggregation level, Andrawis et al. (2011), Kourentzes et al. (2014), Petropoulos and Kourentzes (2014), and Petropoulos and Kourentzes (2015) suggested the use of multiple levels of aggregation, usually abbreviated as MTA (multiple temporal aggregation).”
The inventory performance of forecasting methods: Evidence from the M3 competition data (2019, methodology): “, 2011), a question that still remains is what is the inventory performance of multiple temporal aggregation (Kourentzes et al., 2014)?”
Bagging exponential smoothing methods using STL decomposition and Box–Cox transformation (2016, methodology): “However, in this study we restrict our analysis to the median, as it achieves good results and is less sensitive to outliers than, e.g., the mean, also taking into account the results of Kourentzes et al. (2014a).”
Tales from tails: On the empirical distributions of forecasting errors and their implication to risk (2019, methodology): “Finally, we examine the Bagged ETS (Bergmeir et al., 2016) and the Multi Aggregation Prediction Algorithm (MAPA) (Kourentzes et al., 2014) which exploit bagging and temporal aggregation approaches, respectively, to deal with uncertainty and improve the forecasting accuracy of typical exponential smo”
Neural network ensemble operators for time series forecasting (2014, methodology): “Kourentzes et al. (2014) showed that combining models ﬁtted on data sampled at diﬀerent frequencies can achieve better forecasting accuracy at all short, medium and long term forecast horizons, and found small diﬀerences in using either the mean or the median.”