Analysis of Agricultural Commodities Prices Using BART: A Machine Learning Technique

The price analysis of agricultural commodities has both practical and theoretical aspect. In stark contrast to financial markets, this study purposes applying the Bayesian additive regression tree (BART)—a machine learning technique, for the evaluation and comparison of monthly prices of twenty-five agricultural commodities spanning 2011–2019. The maximum posterior inference function of the BART helps the model evaluate each variable’s marginal effects. The model has the advantage of selecting the intrinsic variable in order to find solution for the problems. The model identifies key predictors and essential interactions among twenty-five independent variables, which are relevant for decision making on the market for agricultural commodities. Overall, the analysis showed that the Bayesian additive regression tree performs significantly better.