Forecasting crude oil prices based on machine learning statistics methods and random sparse Bayesian learning

Abstract

Oil price forecasting has received a great deal of interest from both professionals and scholars because of the unique characteristics of the oil price and its enormous impact on a wide range of economic sectors. In response to this problem, the authors set out to develop a strong model for accurately predicting the Brent crude oil price. We employed the Linear Regression and Random Forest models to examine the market interrelationships present in the oil price time series. Next, the models are given weights such that the experimental time series can be accurately predicted. These errors are quantified in terms of root mean squared errors (RMSE), average errors (MAE), and average percentage errors (MAPE). Results and forecast accuracy of the model as compared to the other model. To maximize their output and order levels and reduce the negative impact of potential shocks, countries that produce and import crude oil benefit greatly from accurate crude oil price forecasts.