波罗的海干散货运价指数预测——基于EMD-XGBoost模型的证据

Prediction of Baltic Dry Bulk Freight Index——Evidence Based on EMD-XGBoost Model

BDI指数是全球经济贸易的晴雨表和大宗商品的风向标,准确预测BDI指数具有重要意义。本文系统选取了金属价格、农产品价格、干散货船舶价格、能源价格以及经济环境等指标,全面刻画BDI指数的影响因素集合。通过EMD分解算法将BDI指数按照频率划分为不同的IMF分量,将分量重构为BDI指数的高频部分、低频部分以及趋势项,利用XGBoost模型对BDI指数进行预测。结果显示:XGBoost模型的可决系数为84%,均方根误差为153.15,相比其他机器学习模型预测效果更为突出,经过EMD算法分解重构后的EMD-XGBoost模型的可决系数达到96%,均方根误差为80.16,效果最优。根据研究结论,得出如下启示:建立航运市场运价的预警体系;构建铁矿石、能源价格与航运运价的联动体系;有效利用各类航运衍生品降低航运市场的运价风险。

The BDI is considered a barometer of global economic and trade conditions and a leading indicator for the direction of bulk commodity prices.Accurate prediction of the BDI index holds significant importance.This article systematically selects indicators such as metal prices,agricultural product prices,dry bulk shipping rates,energy prices,and economic environment to comprehensively depict the factors influencing the BDI index.Through the EMD algorithm,the BDI index is decomposed into different IMFs based on frequency.These IMF components are then reconstructed as the high-frequency component,low-frequency component,and trend component of the BDI index.The XGBoost model is employed to predict the BDI index.The results indicate that the XGBoost model achieves a coefficient of determination(R-squared)of 84%and a root mean square error(RMSE)of 153.15.Compared to other machine learning models,the XGBoost model demonstrates a more prominent predictive performance.After applying the EMD algorithm for decomposition and recon-struction,the EMD-XGBoost model achieves an R-squared of 96%and an RMSE of 80.16,showing the best performance.Based on the research findings,the following insights can be drawn:Establish an early warning system for freight rates in the shipping market;Develop a linkage system between iron ore and energy prices and shipping rates;Effectively utilize various shipping derivatives to mitigate freight rate risks in the shipping market.