关键词: 原油运价指数, 均值回归模型, 时间序列预测, 柯尔莫哥洛夫前向方程

Abstract: In order to reduce the influence of freight fluctuation to the operation and decision-making of crude oil transportation and related enterprises. It is necessary to master the fluctuation law and changing characteristics of international crude oil freight. Effectively grasping the fluctuations and changing characteristics of freight has a vital significance for the operators to invest and manage the companies. In view of Baltic Dirty Tanker Index (BDTI) fluctuation mechanism, and after applying methods of unit root test and variance ratio test to verify the mean reversion characteristic of BDTI verification, this paper analyzes the mean reversion characteristic of BDTI to study its variation trend. Meanwhile, a time related model is established to forecast and analysis the BDTI which improves the traditional mean reversion model, and the corresponding method for parameters estimation is given. The improved model is verified through the numerical simulation both on the fitting effect and prediction accuracy. First, through the characteristics test, the mean regression equation with average level changes over time is established. Second, by using Korlgomov forward differential equation, a prediction model is set up on the base of improved mean reversion model to obtain the predictive value and confidence intervals, then predicted values are obtained through the fitting path of BDTI historical samples. Finally, through the analysis of BDTI published by Clarkson Database, result shows that on the one hand, the crude oil freight index has a significant mean return characteristic. And on the other hand, by forecasting and analyzing the crude oil freight index, the average relative error of this improved model is 0.1597 which is less than the traditional model's 0.1908, while the goodness of fit of improved model is 0.9420 which is a little higher than the tradition's 0.9396. The results show that the improved prediction model has better fitting effect than the traditional model, so it can better reflect the actual situation of the market. This means that the improvement of the mean regression model might be beneficial to the further study and application of the mean regression models. In the aspect of management and operation, the analysis and prediction of freight fluctuation and improvement of the mean regression model will be helpful for investment enterprises to make more excellent investment decision-making proposals.

Key words: tanker freight index, mean reversion model, time series forecasting, Korlgomov forward differential equation