Predicting the price of Vietnamese shrimp products exported to the US market using machine learning

Abstract

Accurately predicting the price of exported fishery products is an important task for fisheries because it will enable market trends to be determined, leading to the development of high-quality fishery products. In this study, we predicted prices in selected base periods (2, 3, 6, and 12 months) to investigate how historical data influenced the Vietnamese export price. A dataset (from May 1995 to May 2019) was collected from the US Department of Agriculture (USDA). We initially hypothesized that the dependent variable, Vietnamese export price, was affected by 33 independent variables, but ultimately used 15 key variables, which were chosen on the basis of Akaike information criterion (AIC) to train the models. A tree-based machine learning technique, including the random forest and gradient boosting tree algorithms, was applied for predictions. It was found that the random forest algorithm performed well for historical data for periods of more than 6 months, while the gradient boosting tree algorithm was better over short durations of less than 6 months.