基于组合模型的油轮运价指数分析与预测  

Analysis and Forecasting of the Tanker Freight Rates Based on Combined Forecasting Model

作　　者：计明军[1] 张海燕[1] 王清斌[1] 

机构地区：[1]大连海事大学交通运输管理学院,辽宁大连116026

出　　处：《交通运输系统工程与信息》2012年第1期199-204,共6页Journal of Transportation Systems Engineering and Information Technology

基　　金：国家自然科学基金(71072081);辽宁省教育厅项目(L20100064);大连市基金资助项目(2009J22DW008);中央高校基本科研业务费专项资金(2011JC009)

摘　　要：分析原油价格对油轮运价指数的影响关系,并预测油轮运价指数发展变化趋势.本文通过Granger因果关联分析,原油价格是油轮运价指数的3阶Granger因.因此,建立了3阶ARCH模型对油轮运价指数进行了预测,预测精度在8%之内.根据油轮运价指数的自身非线性变化趋势,建立了三层BP神经网络模型预测油轮运价指数的发展趋势,精度在3%以内.为进一步提高模型的预测精度,结合ARCH预测模型和BP神经网络预测模型的特点,通过预测误差最小化模型,确定组合权重,建立了新的组合预测模型对未来油轮运价指数进行分析预测,模型的精度控制在2%以内,预测精度显著提高.此研究对油轮运价指数的预测提供了较好的方法.The paper analyzes the relationship between the tanker freight rate and the price of crude oil and forecasts the trend of the tanker freight rates. Using the granger causality test, the price of crude oil is Granger causality of the tanker freight rates with three steps. Therefore, the tanker freight rate is then forecasted by the ARCH model with three steps. The accuracy of the model is within 8%. Based on the nonlinear trend of tanker freight rates, the BP neural network with three levels is used to forecast the tanker freight rates. The accuracy of the model is within 3%. Combining the characteristics of the ARCH model and the BP neural network, a novel combined forecasting model is modified to improve the accuracy of tanker freight rate forecast, whose weights are gained by minimizing the model of forecasting errors. The accuracy of the model is within 2% and significantly improved. The study provides a good method for the tanker freight rate forecast.

关 键 词：水路运输 组合预测模型 ARCH模型 BP神经网络模型 油轮运价指数 

分 类 号：U695[交通运输工程—港口、海岸及近海工程]