A Blending Ensemble Learning Model for Crude Oil Price Prediction

Given that the price of crude oil is driven by a number of factors with varying frequency, it is difficult to accurately capture its behavior, which in turn leads to challenges in forecasting. Moreover, different mechanisms of fluctuations have been observed at different time series periods. To efficiently capture these diverse fluctuation profiles, we propose to combine heterogenous predictors for predicting the crude oil price. Specifically, a forecasting model is developed using blended ensemble learning is developed that combines various machine learning methods, including linear regression, k-nearest neighbor regression, regression trees, support vector regression, and ridge regression. Brent and WTI crude oil data at various time series frequencies are used to validate the proposed blending ensemble learning approach. To show the effectiveness of the proposed model, its performance is compared with existing individual and ensemble learning methods used for crude oil price prediction, such as lasso regression, bagging lasso regression, boosting, random forest, and support vector regression. We show that our proposed blending ensemble learning model dominates the existing forecasting models in terms of forecasting errors. The proposed model exhibits a good prediction performance for both short- and long-term forecasting horizons, which is beneficial to stakeholders and related industries that depend on this energy source