A Machine Learning Approach to Forecast International Trade: The Case of Croatia

Abstract Background: This paper presents a machine learning approach to forecast Croatia’s international bilateral trade. Objectives: The goal of this is evaluate the performance algorithms in predicting trade flows related imports and exports case Croatia. Methods/Approach: dataset on Croatian with over 180 countries worldwide from 2001 2019 assembled using main variables gravity model. To values for horizon one year (the 2020), (Gaussian processes, Linear regression, Multilayer perceptron) have been used. Each forecasting algorithm evaluated by calculating mean absolute percentage errors (MAPE). Results: It was found that very good ability trade, neural network perceptron having best among other algorithms. Conclusions Main findings can be important economic policymakers subjects field research. Timely information about changes trends projections future significantly affect decision-making flows.