Forecasting Domestic Shipping Demand of Cement: Comparison of SARIMAX, ANN and Hybrid SARIMAX-ANN

Abstract

Accurate forecasting of shipping demand for cement is valuable for the ensure of proper supply and demand match. Due to the perishability nature of the shipping industry, accuracy performances of the models in the industry is particularly important. Therefore, supply and demand in the industry should match. This paper compares three methods, Seasonal Autoregressise Integrated Moving Average with Exogenous Variable (SARIMAX), Artificial Neural Network-Multilayer Perceptron (MLP), and SARIMAX-MLP hybrid model for domestic shipping demand of cement loaded at ports of Turkey. The hybrid model forecasting of the shipping demand of cement appears to be more applicable than that of the single SARIMAX. The results indicate that the prediction accuracy of SARIMAX model is higher than the MLP model and that the hybrid model is the most satisfactory of the three models.