Directional Forecasting of WTI and Brent Crude Oil Prices: A Machine Learning Approach with Technical Indicators at Daily, Weekly, and Monthly Frequencies

Crude oil prices exhibit pronounced volatility, nonstationarity, and nonlinear behavior, making accurate forecasting inherently challenging, particularly when employing traditional statistical models such as the Autoregressive Integrated Moving Average (ARIMA) model. Although classical time-series techniques remain widely adopted by market practitioners, technical indicators have received comparatively limited attention in the academic literature on energy price forecasting. To address this research gap, the present study employs supervised machine learning algorithms—including Support Vector Machine (SVM), Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), Decision Tree (DT), Naïve Bayes (NB), and Random Forest (RF)—to forecast the directional movement (up or down) of two major crude oil benchmarks, West Texas Intermediate (WTI) and Brent, across three temporal frequencies: daily, weekly, and monthly, for the period 2010 to 2024. While these algorithms are capable of solving both regression and classification problems, this research specifically formulates crude oil price forecasting as a binary classification task, wherein the target variable indicates whether the price will rise or decline in the subsequent time interval. Model performance is assessed using four widely accepted classification metrics: accuracy, precision, recall, and F1-score. Empirical results demonstrate that SVM models—particularly those employing linear and polynomial kernels—consistently achieve superior forecasting accuracy compared to other classifiers across most experimental settings.