Forecasting the Crude Oil Spot Price with Bayesian Symbolic Regression

In this study, the crude oil spot price is forecast using Bayesian symbolic regression (BSR). particular, initial parameters specification of BSR analysed. Contrary to conventional approach regression, which based on genetic programming methods, applies algorithms evolve set expressions (functions). This econometric method able deal with variable uncertainty (feature selection) issues in forecasting. Secondly, research seems be first application Monthly data between January 1986 and April 2021 are As well as BSR, several other methods (also uncertainty) used benchmark models, such LASSO ridge regressions, dynamic model averaging, averaging. The more common ARIMA naïve also used, together time-varying parameter regressions. a result, not only presents novel original but provides concise uniform comparison popular forecasting for price. Robustness checks performed strengthen obtained conclusions. It found that suitable selection functions operators initialization an important, trivial, task. Unfortunately, does result forecasts statistically significantly accurate than models. However, computationally faster programming-based regression.