Multistep Ahead Forecasting of WTI Crude Oil Prices Using Time Series and Machine Learning Models

Crude oil is a critical energy source and a raw material for various products, including fuels (gasoline, diesel, and jet fuel), lubricants, petrochemicals, and asphalt. It undergoes refining processes to separate and convert it into usable products. Multiple factors, including economic indicators, geopolitical events, supply and demand dynamics, and technological advancements influence crude oil prices. Its global market has prices often benchmarked to grades like Brent Crude and West Texas Intermediate (WTI). In this article, the daily WTI crude oil prices are multi-step ahead forecasted using four different methods which are autoregressive integrated moving average (ARIMA), random forest (RF), k-nearest neighbors (kNN), and autoregressive neural networks (ARNN). The data set consists of daily WTI crude oil prices from February 13, 2014, to February 13, 2024, divided into 80% training set and 20% validation set. The performance of the methods is evaluated by one-step and seven-step ahead forecasting using root mean square error (RMSE) and mean absolute percentage error (MAPE) as the accuracy measures. The results show that ARIMA outperforms the methods for one and seven-step ahead forecasting, followed by ARNN, kNN, and RF. The study provides useful insights for investors and policymakers in the oil market.

Hali tarjima qilinmagan