Distributed Lag Non-linear Model을 이용한 Baltic Dry Index 예측 및 요인 별 시차 영향도 도출에 관한 연구

Abstract

The Baltic Dry Index (BDI) is a crucial and representative indicator in the shipping logistics industry sector for evaluating the market conditions, providing valuable information on the movement of global trade and production operations. The BDI, known for its significant volatility, has been the focus of several research initiatives utilizing machine learning and deep learning approaches to improve accuracy. Nevertheless, these approaches frequently encounter difficulties in understanding the results of the predictions. This study provides a framework that utilizes the Distributed Lag Non-linear Model to enhance the predictive accuracy of the BDI. The proposed model has the capability of capturing lag effects based on different factors to improve the prediction results. To evaluate the efficacy of the suggested approach, we carried out comparison experiments utilizing four performance metrics that have been widely utilized in previous studies to compare with different machine learning and deep learning models. The results showed enhanced prediction accuracy in comparison with existing methodologies. The lag effects by factor have been examined using the trained DLNM model, and we verified that the analyzed results are in good agreement with prior research by comparing them with previous research