A model to estimate the origin–destination matrix for soybean exportation in Brazil

This article proposes a model to estimate the origin–destination (O–D) matrix for the export flow of Brazilian soybeans based upon a constrained gravity model. The performance of the model was improved by establishing an adjustment criterion for the data, which increased the coefficient of determination of the trend by 24–46%, combined with multi-objective optimization techniques to ensure a balance of the O–D matrix, which yielded R2 values of 0.94 regardless of the prediction year. The model is easy to apply and can help in the strategic planning of transportation for the export of soy products from Brazil, provided that the supply chain configuration is known and the absolute percentage errors are monitored over time.

► We model the origin–destination matrix for Brazil’s export flows of Brazilian soybean.
► We employ a constrained gravity model.
► We employ multi-objective optimization techniques to assure the balance of the O–D matrix.
► The model enabled attaining R2 values of 0.94.