Multivariate analysis and forecasting of the crude oil prices: Part I – Classical machine learning approaches

Crude oil price fluctuations have extensive consequences for economic growth, inflation rates, industrial production, and supply chains. Accurate prediction of crude oil prices enables proactive decision-making and risk management, making it a vital area of research for policymakers and industry leaders. A widespread notion among relevant studies on oil price prediction is the predominant implementation of time series and traditional econometric models. This research aims to narrow the divide between utilizing multiple variables and departing from conventional econometric models for anticipatory analysis of oil prices. In this study, we utilized multivariate analysis to forecast crude oil spot prices using Support Vector Regression (SVR). Multivariate regression models with and without subset selection, Ridge regression, and LASSO regression as comparable models are also developed and analyzed. Our findings demonstrate that the SVR model with radial basis function (RBF) kernel offers superior forecasting accuracy and greater stability than other models with the very least RMSE value of 0.0766. Furthermore, this model has good generalization capabilities and an accurate representation of the time-variance and nonlinearity. Moreover, our comparative analysis demonstrates that the SVR model with RBF kernel surpasses widely well-known techniques in crude oil price prediction by exhibiting superior accuracy in directional and level predictions.