Institute of Transportation, MOTC

     International trade mainly relies on maritime transportation. The Baltic Dry Index (BDI) is an important indicator representing the freight rate of the dry bulk shipping market. However, analysis of this trend and the seasonal fluctuation patterns of the BDI have long been considered challenging. This research seeks to predict these elements by combining traditional time series model and data mining techniques into the hybrid model. Specifically, this work employs Seasonal and Trend decomposition using Loess (STL) to decompose the time series data into three components: the trend, seasonality, and reminders. The trend is predicted by an Autoregressive Integrated Moving Average (ARIMA) model, while the remainder is predicted by Classification And Regression Tree (CART) model. These two techniques were adopted because the ARIMA model is useful for trend prediction and the CART performs well when predictingnon-linear time series data. BDI data collected between 1998 and 2016 was used to train and test the performance of the proposed method. The results show that combining the CART and ARIMA models enhances predictive performance not only with regards to the trend, but also to the fluctuations of the BDI. This was verified by the mean absolute percentage error, which showed that the overall prediction effect was good. The proposed method will provide a strategic analysis tool for dry bulk shipping carriers that can be used as a part of their greater decision-support system