Search

The COVID-19 outbreak and the implementation of peak and carbon neutral policies have severely impacted oil price volatility and the industrial sector. Exploring the impact mechanisms between oil prices and industries is particularly important for accurate forecasting of crude oil prices. As one of the world's largest commodity consumers, China's crude oil market is more representative and susceptible to external factors than that of developed countries. In this paper, we propose an analytical forecasting framework based on the causal effects between Shanghai crude oil prices and various industries in China to improve the forecasting accuracy of crude oil prices. Information geometric causal inference (IGCI) is applied to detect causal relationships between 31 different industries in China and Shanghai crude oil prices in the three time periods before, during and after COVID-19, and industries with strong causal information effects on crude oil prices in the long run are screened out as additional features. An oil price forecasting model based on Gray Wolf Optimization and Extreme Learning Machine (GWO-ELM) is proposed. Considering the small amount of data for Shanghai crude oil, this paper proposes a cross-learning data approach to solve the problem. Experimental results show that the GWO-ELM model outperforms RF, LSTM, GRU, and migration learning-based Tr-LSTM and Tr-Adaboost models in the task of Shanghai crude oil futures price prediction, and find that industry characteristics with long-term causal effects on oil prices can improve the model prediction accuracy. Our proposed analytical prediction can capture the oil price trend more accurately through the information of the industry and solve the problem of insufficient training data for the model. The application of this framework is expected to provide new methods and ideas for data mining of crude oil and other futures prices. [ABSTRACT FROM AUTHOR]