Demand Forecasting for Freight Transport Applying Machine Learning into the Logistic Distribution

The use of information technologies such as Transportation Management System (TMS) is crucial to improve the transportation process in a company, offering potential results that include optimization of shipments, an increase in cost savings and customer satisfaction. The TMS, along with data analysis techniques such as Machine Learning (ML) and/or Data Mining (DM) are useful tools for demand forecasting, facilitating strategic, tactical and operational decision making, which ultimately improves the logistics distribution performance. For this, this work proposes a methodology that will compare three prediction methods (traditional statistical method, hybrid method and artificial neural network), which will be tested through the Mean Squared Error (MSE) indicator to determine which one has more accurate results in the forecast of demand for freight transport. The experimental results obtained in this research show that the historical data saved in a TMS can be used for reliably freight transportation demand forecasting. Each of the three methods proposed in this paper has been an effective tool in resolving time series forecasting problems, and when Artificial Neuronal Network (ANN) was used, better results are yielded.