Forecasting selected energy commodities prices with Bayesian dynamic finite mixtures

摘要: Abstract In this paper selected energy commodities spot prices are forecasted with the help of Bayesian dynamic finite mixtures. particular, crude oil, natural gas, and coal analyzed. Due to availability data, oil is analyzed between 1988 2019, gas 1990 1987 2019. Monthly data used. The mixtures used herein a novel methodological tool in forecasting. Their first important feature that regression coefficients estimated recursive on-line way, allowing for real-time performance. Secondly, switching mixture components also allowed vary time. Thirdly, algorithms based on explicit solutions, fully inference approach, whereas approximations only numerical level pdfs (probability density functions) statistics. other words, evolution prior posterior has fixed functional form; statistics those evolving Both normal state-space models considered as components, which makes study generalization previous research approaches model averaging techniques. Indeed, compared benchmark models, such Dynamic Model Averaging, Time-Varying Parameter regression, ARIMA, naive method, Diebold-Mariano test, found generate significantly more accurate forecasts. Additionally, Giacomini-Rossi fluctuation test Confidence Set applied thorough examination forecasting performances.