Forecasting crude oil option prices with dynamic factors using integrated machine learning models

In this article, we develop a dynamic factor model utilizing the incremental principal component analysis (IPCA) method to identify the key factors influencing crude oil option prices. We employ eight popular machine learning algorithms, i.e., LASSO, Support Vector Regression (SVR), Ridge Regression, Back Propagation (BP), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Multi-Layer Perceptron (MLP), to forecast crude oil prices. Their performance is rigorously evaluated using four distinct error measurements alongside the model confidence set (MCS) and direction of change (DoC) tests. We further introduce an integrated machine learning model that combines the strengths of the top-performing algorithms through weighted averaging, enhancing both predictive accuracy and robustness. In addition, we also explore the interpretability of the factors influencing the precision of predictivity using the SHAP value, so that the contribution of these factors towards the prediction is clear. Our findings offer substantial potential for mitigating energy risks and improving market efficiency through refined price forecasting techniques.