科研管理

科研管理 ›› 2017, Vol. 38 ›› Issue (5): 150-160.

• 论文 • 上一篇    

综合周期、均值回复和跳跃特性的BDI指数预测研究

余方平,匡海波   

  1. 大连海事大学综合交通运输协同创新中心，辽宁 大连116026
  • 收稿日期:2017-02-07 修回日期:2017-04-01 出版日期:2017-05-20 发布日期:2017-05-20
  • 通讯作者: 余方平
  • 基金资助:

    国家自然科学基金：自贸区港口生态圈演化、平衡及评价机制研究（71672016）；长江学者和创新团队发展计划资助：港口协同发展与绿色增长（IRT13048）；河北省交通运输厅重点项目：河北港群绿色增长发展模式研究（ZJT2015037）。

Research on forecasting BDI index with comprehensive periodicity, mean reversion and jump features

Yu Fangping, Kuang Haibo   

  1. Collaborative Innovation Center for Transport Studies, Dalian Maritime University, Dalian 116026, Liaoning, China
  • Received:2017-02-07 Revised:2017-04-01 Online:2017-05-20 Published:2017-05-20

摘要： BDI指数的预测对于航运市场经营管理具有重要地指导作用。本研究构建了基于周期、均值回复和跳跃特性的BDI指数O-U随机预测模型，主要创新点有：一是分析了BDI指数的周期、均值回复以及跳跃特性，将该三大特性纳入BDI指数随机预测模型，有效提升了BDI指数预测理论科学性。二是借助O-U随机过程，建立了基于周期、均值回复以及跳跃特性BDI指数预测模型，同时，利用Fourier级数函数估计周期参数，借助一阶自回归估计均值回复参数，以及Gamma分布和双指数分布来估计跳跃参数，解决了参数较多、估计难度较大的问题。三是采集2013年-2015年BDI指数日数据进行拟合，并借助蒙特卡罗方法对2016年上半年BDI指数开展了预测，结果表明本模型预测精确度较高。

关键词: BDI指数, 周期, 均值回复, 跳跃, O-U随机过程, 预测

Abstract: BDI index forecasting is of great practical significance for the management and decision-making of shipping market. In this study, we construct a O-U stochastic forecasting model of BDI index based on A new perspective from periodicity, mean reversion and jump features in this paper, The main innovation points are: Firstly, The periodicity, mean reversion and jump features of BDI index are analyzed and included in the BDI Index Stochastic forecasting model, which effectively improves the accuracy of the BDI index forecasting theory. Secondly, the BDI index stochastic forecasting model based on periodicity, mean reversion and jump features is set up. Simultaneously, the periodic parameters are estimated by the Fourier series function, the mean reverting parameters are estimated by means of the first order autoregressive estimation, and the jump parameters are estimated by means of Gamma distribution and double exponential distribution, which solves the problem of more parameters extremely difficult to calculate. Thirdly, the daily BDI data in 2013 to 2015 are presented to test the proposed model. The BDI index was fitted, and in the first half of 2016 is forecasting by Monte Carlo method. The results show that the proposed method has higher accuracy.

Key words: BDI index, cycle, mean reversion, jump, O-U stochastic process, forecasting