Artificial Intelligence Meets Computational Intelligence : Multi-Commodity Price Volatility Accuracy Forecast with Variants of Markov-Switching-GARCH-Type-Extreme Learning Machines Hybridization Framework

The price volatility of energy assets such as natural gas, crude oil, and coal among others do influence electricity prices, which altogether directly have significant economic impacts on different sectors of the economy. From this viewpoint, accurate energy price volatility predictions are very valuable for the reliable and stable operational security of energy systems. However, forecasting energy price volatility can be less satisfactory, especially in the era of big data because of the complexity and non-linearity inherent in the energy system. This paper, therefore, utilizes a hybridization of techniques that seamlessly cut across artificial intelligence and computational intelligence for forecasting energy price volatility. In particular, we propose three parallel forecast combinations of Markov Switching GARCH-type and extreme learning machine model (MS-GARCH-type-ELM) alongside the selection of an appropriate model for price volatility forecasting in the energy market. Using energy time series data on coal, gas, and crude oil, we examine the forecasting performance of the different models. Our findings show that the Mean Absolute Error (MAE), Mean Square Error (MSE), Theil's U and Mean Absolute Percentage Error (MAPE) indicate that the MS-GARCH-type-ELM model with a simple structure could obtain more accurate forecasting results and as such the best forecasting results at all steps among all related models and hence, can help to predict short-run price volatility fluctuations in the energy market. In addition, investigating both the Pre-COVID and during COVID outbreak periods reveal that OLS-MSGARCH-tGARCH-GJR-GARCH-ELM with combined student-t and skewed student-t distribution is robust and show resilience as competing combined accurate forecast model for crude oil price volatility forecasting while the remaining show mixed results in the two paradigms namely (i) Pre-COVID period and (ii) during the COVID-19 outbreak. Furthermore, we employ the model confident set (MCS) to check the superiority of our models, which reveals that all the models including the benchmark favor the different commodity assets in the COVID era, however; the opposite situation is observed in the non-COVID era. Overall, the (MS-GARCH-type-ELM) models are capable of providing better accurate forecasts in stressful times compared to non-stressful periods