A Deep Learning Framework for Baltic Dry Index Forecasting
Mingxi Liu, Yajie Zhao, Jingkai Wang, Chang Liu, Guowen Li
Abstract
Baltic Dry Index (BDI) is a widely used index that measures the changes in the transportation freight rates of the dry bulk shipping market, which could also serve as an indicator of the prosperity of both general shipping market and the world economy. As a typical non-stationary and non-linear time series, BDI is influenced by a variety of complicated factors, and thus accurately forecasting BDI is undoubtedly an important and challenging task. In this paper, a deep learning based framework for BDI forecasting is proposed, which contains a deep learning based feature extraction process and a machine learning regression process. Compared with traditional econometric and machine learning methods, this framework could greatly boost the prediction accuracy through capturing some in-depth features.