Crude Oil Price Prediction Based on Multiple Ensemble Learning Algorithms

Abstract

The volatility of crude oil price affects people from all walks of life. Since the start of the 21st century, the frequent booms and busts in crude oil price have caused a greater emphasis on forecasting the price. In order to make prediction from a new perspective, this paper combined both financial and non-financial factors and applied both bagging and boosting algorithms in ensemble learning to construct models. According to the results, both bagging and boosting algorithms achieved impressive improvements over the benchmark model. The ensemble learning models can not only reach high goodness-of-fit, but also attain excellent accuracy in direction prediction, which indicates the models can well realize the long term trend prediction of crude oil price. In brief, this paper enriched the content in predicting crude oil price and offer a guideline for the application of ensemble learning algorithms in such a field.