

The Baltic Dry Index (BDI) is an indicator of freight rates of dry bulk cargo. Since freight rate fluctuates widely with reflecting a rapid change in the shipping market, it is important for a shipping company to predict the BDI. In the shipping industry, detailed data on the movement of each vessel is currently available in the automated identification system (AIS) on board, and a wide use of the AIS data is increasingly expected. This study proposes a new prediction method for freight index using deep learning and AIS data. The prediction target is the Baltic Capesize Index (BCI) representing the freight rate index of a large dry-bulk carrier over 180,000 DWT. The AIS data and various statistics are incorporated into the model to predict the rise or fall of the BCI value after 30 days. A multiple set of AIS data of the entire world and several specific regions are used. Furthermore, the number of related statistics to be incorporated is increased and a method of selecting them by introducing maximal information coefficient is shown. From the simulation results, the prediction of BCI can be performed with a certain degree of accuracy. In addition, the effect of introducing AIS data in the BCI prediction is confirmed.