Forecasting the containerized freight index with AIS data: A novel information combination method based on gray incidence analysis

This paper uses the container shipping capacities of 11 major trade lanes, obtained from automatic identification system (AIS), to construct a common factor based on gray incidence analysis (GIA) in the aim of improving the predictability of containerized freight index. Our results show that the common factor generated by GIA consistently exhibits better out‐of‐sample prediction performances than principal component analysis (PCA) and least absolute shrinkage and selection operator (LASSO), meaning that GIA can extract more useful information for forecasting freight index. Our main findings are first, GIA can evaluate the similarity between the predictors and the predicted value. Unlike popular information combination method PCA, which cannot extract the relevant information from the predictors, GIA can extract the most relevant information of the predictors to the predicted value. Second, different from LASSO, which drops some information, GIA maintains the most information, because the container shipping capacities of different lanes all impact the freight index. Third, AIS data do provide information increments for freight rate forecasting. This research explores a new field application of gray relational analysis in information combination and presents one application of GIA in big data processing. This research shows the usefulness of AIS information in predicting freight index. Additionally, this research enlightens the prediction of freight rate based on big data from AIS.