nep-for New Economics Papers
on Forecasting
Issue of 2021‒10‒25
six papers chosen by
Rob J Hyndman
Monash University

  1. Machine learning in energy forecasts with an application to high frequency electricity consumption data By Erik Heilmann; Janosch Henze; Heike Wetzel
  2. Macroeconomic Forecasting with LSTM and Mixed Frequency Time Series Data By Sarun Kamolthip
  3. Conditional Heteroscedasticity Models with Time-Varying Parameters: Estimation and Asymptotics By Armin Pourkhanali; Jonathan Keith; Xibin Zhang
  4. Adaptive Learning on Time Series: Method and Financial Applications By Parley Ruogu Yang; Ryan Lucas; Camilla Schelpe
  5. Robust estimation and forecasting of climate change using score-driven ice-age models By Blazsek, Szabolcs Istvan; Escribano Sáez, Álvaro
  6. Erratum to 'Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark' [Appl. Energy 293 (2021) 116983] By Jesus Lago; Grzegorz Marcjasz; Bart De Schutter; Rafal Weron

  1. By: Erik Heilmann (University of Kassel); Janosch Henze (University of Kassel); Heike Wetzel (University of Kassel)
    Abstract: Forecasting plays an essential role in energy economics. With new challenges and use cases in the energy system, forecasts have to meet more complex requirements, such as increasing temporal and spatial resolution of data. The concept of machine learning can meet these requirements by providing different model approaches and a standardized process of model selection. This paper provides a concise and comprehensible introduction to the topic by discussing the concept of machine learning in the context of energy economics and presenting an exemplary application to electricity load data. For this, we introduce and demonstrate the structured machine learning process containing the preparation, model selection and test of forecast models. This process is intended to serve as a general guideline for energy economists and practitioners who need to apply sophisticated forecast models.
    Keywords: machine learning, electricity consumption forecast, artificial neural network, time series forecast
    JEL: C45 C53 Q47
    Date: 2021
  2. By: Sarun Kamolthip
    Abstract: This paper investigates the potentials of the long short-term memory (LSTM) when applying with macroeconomic time series data sampled at different frequencies. We first present how the conventional LSTM model can be adapted to the time series observed at mixed frequencies when the same mismatch ratio is applied for all pairs of low-frequency output and higher-frequency variable. To generalize the LSTM to the case of multiple mismatch ratios, we adopt the unrestricted Mixed Data Sampling (U-MIDAS) scheme (Foroni et al., 2015) into the LSTM architecture. We assess via both Monte Carlo simulations and empirical application the out-of-sample predictive performance. Our proposed models outperform the restricted MIDAS model even in a set up favorable to the MIDAS estimator. For real world application, we study forecasting a quarterly growth rate of Thai real GDP using a vast array of macroeconomic indicators both quarterly and monthly. Our LSTM with U-MIDAS scheme easily beats the simple benchmark AR(1) model at all horizons, but outperforms the strong benchmark univariate LSTM only at one and six months ahead. Nonetheless, we find that our proposed model could be very helpful in the period of large economic downturns for short-term forecast. Simulation and empirical results seem to support the use of our proposed LSTM with U-MIDAS scheme to nowcasting application.
    Keywords: LSTM; Mixed Frequency Data; Nowcasting; Time Series; Macroeconomic Indicators
    JEL: E37 C35
    Date: 2021–10
  3. By: Armin Pourkhanali; Jonathan Keith; Xibin Zhang
    Abstract: This paper proposes using Chebyshev polynomials to approximate time-varying parameters of a GARCH model, where polynomial coefficients are estimated via numerical optimization using the function gradient descent method. We investigate the asymptotic properties of the estimates of polynomial coefficients and the subsequent estimate of conditional variance. Monte Carlo studies are conducted to examine the performance of the proposed polynomial approximation. With empirical studies of modelling daily returns of the US 30-year T-bond daily closing price and daily returns of the gold futures closing price, we find that in terms of in-sample fitting and out-of-sample forecasting, our proposed time-varying model outperforms the constant-parameter counterpart and a benchmark time-varying model.
    Keywords: : Chebyshev polynomials, function gradient descent algorithm, loss function, one-day-ahead forecast
    JEL: C14 C58
    Date: 2021
  4. By: Parley Ruogu Yang; Ryan Lucas; Camilla Schelpe
    Abstract: We formally introduce a time series statistical learning method, called Adaptive Learning, capable of handling model selection, out-of-sample forecasting and interpretation in a noisy environment. Through simulation studies we demonstrate that the method can outperform traditional model selection techniques such as AIC and BIC in the presence of regime-switching, as well as facilitating window size determination when the Data Generating Process is time-varying. Empirically, we use the method to forecast S&P 500 returns across multiple forecast horizons, employing information from the VIX Curve and the Yield Curve. We find that Adaptive Learning models are generally on par with, if not better than, the best of the parametric models a posteriori, evaluated in terms of MSE, while also outperforming under cross validation. We present a financial application of the learning results and an interpretation of the learning regime during the 2020 market crash. These studies can be extended in both a statistical direction and in terms of financial applications.
    Date: 2021–10
  5. By: Blazsek, Szabolcs Istvan; Escribano Sáez, Álvaro
    Abstract: ScScore-driven models applied to finance and economics have attracted significant attention in the last decade. In this paper, we apply those models to climate data. We study the robustness of a recent climate econometric model, named ice-age model, and we extend that model by using score-driven filters in the measurement and transition equations. The climate variables considered are Antarctic ice volume Icet, atmospheric carbon dioxide level CO2,t, and land surface temperature Tempt, which during the history of the Earth were driven by exogenous variables. The influence of humanity on climate started approximately 10-15 thousand years ago, and it has significantly increased since then. We forecast the climate variables for the last 100 thousand years, by using data for the period of 798 thousand years ago to 101 thousand years ago for which humanity did not influence the Earth’s climate. For the last 10-15 thousand years of the forecasting period, we find that: (i) the forecasts of Icet are above the observed Icet, (ii) the forecasts of the CO2,t level are below the observed CO2,t, and (iii) the forecasts of Tempt are below the observed Tempt. Our results are robust, and they disentangle the effects of humanity and orbital variables.
    Keywords: Climate Change; Dynamic Conditional Score Models; Generalized Autoregressive Score Models; Ice-Ages and Inter-Glacial Periods; Atmospheric Co2 and Land Surface Temperature
    Date: 2021–10–14
  6. By: Jesus Lago; Grzegorz Marcjasz; Bart De Schutter; Rafal Weron
    Abstract: This Erratum corrects the error metrics of the LEAR models for the German (EPEX DE) market reported in Tables 2 and 3 of Lago et al. (2021) Applied Energy 293, 116983.
    Keywords: Electricity price forecasting; Regression model; Deep learning; Open-access benchmark; Forecast evaluation; Best practices
    JEL: C22 C32 C45 C51 C53 Q41 Q47
    Date: 2021–07–08

This nep-for issue is ©2021 by Rob J Hyndman. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.