A novel approach to Predict WTI crude spot oil price: LSTM-based feature extraction with Xgboost Regressor

Abstract

This paper presents a novel model based on LSTM to predict future prices of WTI crude oil. The WTI price forecasting utilizes data on spot gold price, US 10-year bond yield, global economic activity, and US dollar index from January 1986 to May 2023. The model's performance is assessed using measures such as MAE, MSE, RMSE, MAPE, and R2 metrics. The results generated by the proposed new model are compared to those of the existing machine and deep learning methods, and it is observed that the new model performs better than the existing models in all statistical tests. The study further examined the decision-making processes of the model using SHAP analysis and assessed the individual contribution of each feature to the model's predictions. The correlation between the US Dollar Index and Gold prices and WTI crude oil prices is evident. The SHAP research has demonstrated that the model effectively captures complicated economic linkages and enhances the accuracy of forecasts. The results of this study enhance the development of models that are capable of predicting results, even in times of significant instability, such as economic crises. Using sophisticated data analytics and AI methods would improve the efficiency of energy market oversight.