|
on Forecasting |
| By: | Katarzyna Chec; Bartosz Uniejewski; Rafal Weron |
| Abstract: | Recent studies provide evidence that decomposing the electricity price into the long-term seasonal component (LTSC) and the "residual", predicting both separately, and then combining their forecasts can bring significant accuracy gains in day-ahead electricity price forecasting. However, not much attention has been paid to predicting the LTSC, and the last 24 hourly values of the estimated pattern are typically copied for the target day. To address this gap, we introduce a novel approach which extracts the trend-seasonal pattern from a price series extrapolated using price forecasts for the next 24 hours. Analyzing data from the German and Spanish markets, and considering parsimonious autoregressive and LASSO-estimated models, we find that improvements in predictive accuracy can be as high as 16% over a 5-year test period covering the Covid-19 pandemic, the 2021/2022 energy crisis, and the war in Ukraine. |
| Keywords: | Electricity price forecasting; Long-term seasonal component; Day-ahead market; Combining forecasts |
| JEL: | C22 C51 C53 Q41 Q47 |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:ahh:wpaper:worms2404 |
| By: | Parley R Yang; Alexander Y Shestopaloff |
| Abstract: | We demonstrate the use of Conditional Variational Encoder (CVAE) to improve the forecasts of daily stock volume time series in both short and long term forecasting tasks, with the use of advanced information of input variables such as rebalancing dates. CVAE generates non-linear time series as out-of-sample forecasts, which have better accuracy and closer fit of correlation to the actual data, compared to traditional linear models. These generative forecasts can also be used for scenario generation, which aids interpretation. We further discuss correlations in non-stationary time series and other potential extensions from the CVAE forecasts. |
| Date: | 2024–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2406.19414 |
| By: | Yoosoon Chang (Department of Economics, Indiana University); Yong-gun Kim (Bank of Korea); Boreum Kwak (Bank of Korea); Joon Y. Park (Department of Economics, Indiana University) |
| Abstract: | In this paper, we study how we may use density forecasts to improve point forecasts for the Korean GDP growth rates during the period from 2013:Q3 to 2022:Q1. Although the time span under investigation is much shorter than desired, our conclusions are clear. Density forecasts improve point forecasts, as long as they are effectively approximated and represented as finite dimensional vectors by appropriately chosen functional bases. However, they may only be used to adjust point forecasts. Combining them with point forecasts to define weighted mean forecasts does not yield any meaningful improvement. The functional bases we use for our baseline approach are the leading functional principal components, which by construction most efficiently extract the variations in density forecasts over time. To disentangle the effects of the mean and other aspects of density forecasts, however, we also use the functional basis, which designates, as the leading factor, the mean factor that captures the temporal changes in the mean of density forecasts. Especially with the use of this functional basis, we see a drastic increase in the precision of point forecasts for the Korean GDP growth rates. In fact, the mean squared error of point forecasts decreases by more than 33%, if they are adjusted by density forecasts with our functional basis including the mean factor. |
| Keywords: | GDP growth rate, point forecast, growth-at-risk density forecast, functional regression, functional basis, functional principal component analysis |
| Date: | 2024–05 |
| URL: | https://d.repec.org/n?u=RePEc:inu:caeprp:2024005 |
| By: | Liang, Weifang; Liu, Yong; Somogyi, Simon; Anderson, David P. |
| Keywords: | Research Methods/Statistical Methods |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:ags:aaea22:343687 |
| By: | Andrei Renatovich Batyrov |
| Abstract: | There are several approaches to modeling and forecasting time series as applied to prices of commodities and financial assets. One of the approaches is to model the price as a non-stationary time series process with heteroscedastic volatility (variance of price). The goal of the research is to generate probabilistic forecasts of day-ahead electricity prices in a spot marker employing stochastic volatility models. A typical stochastic volatility model - that treats the volatility as a latent stochastic process in discrete time - is explored first. Then the research focuses on enriching the baseline model by introducing several exogenous regressors. A better fitting model - as compared to the baseline model - is derived as a result of the research. Out-of-sample forecasts confirm the applicability and robustness of the enriched model. This model may be used in financial derivative instruments for hedging the risk associated with electricity trading. Keywords: Electricity spot prices forecasting, Stochastic volatility, Exogenous regressors, Autoregression, Bayesian inference, Stan |
| Date: | 2024–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2406.19405 |