Rapid technological improvements in data digitization and computing power have paved the way for more advanced quantitative techniques in market analysis, particularly in the area of machine learning. This study utilized two traditional econometric and four machine learning techniques for a side-by-side comparison of their effectiveness in short-term forecasting of international corn basis in the five major global export markets. The models were developed under two different forecasting regimes representing a structural break brought on by the COVID-19 pandemic and related concurrent events with the latter regime characterized by a significant increase in volatility. Machine learning offered considerable improvement in out-ofsample forecast performance measurements when compared to econometric methods, particularly in the latter, more volatile forecasting regime. An analysis of the feature selection and variable rankings indicated substantial diversity of selection across the modeling techniques; however, some common observations were derived from the results.
