基于SVM和混沌时间序列的干散货运价指数预测研究

发布时间：2018-11-21 20:33

【摘要】：作为干散货航运市场的“晴雨表”,干散货运价指数反映了干散货运输市场的运价水平。由于受到多种因素的影响,近年来干散货运价指数始终处于剧烈波动之中,且走势难以琢磨,表现出了复杂的非线性特征,传统的预测方法难以取得良好的预测效果,这也给干散货航运市场经营者的决策带来了困难。干散货运价指数波动剧烈,蕴含了国际干散货航运市场长期以来的演化信息。本文在深刻分析干散货运价指数波动的内在规律及外在影响的基础上,提出结合混沌时间序列分析和支持向量机(Support Vector Machine, SVM)回归原理的混合预测模型,对干散货运价指数(Baltic Dry Index, BDI)进行了有效地预测。本文首先对国际干散货航运的供需市场进行深入分析,揭示了干散货市场运价波动的内在原因。其次,文中阐述了干散货运价指数的成因及航线构成,并对运价指数的影响因素及波动性进行了定性分析,为选择适当的预测方法奠定了基础。鉴于干散货运价指数的非线性特征,本文提出了结合混沌时间序列分析的相空间重构和支持向量机(SVM)的混合预测模型,探讨并阐述了混合模型的预测原理及建模思路。接着,本文在对混合预测模型关键参数的选取进行系统分析的基础上,建立了参数联合优化问题的数学模型,并采用遗传算法对该优化问题进行求解。最后,选取BDl月度均值进行实证分析,对BDI样本序列进行混沌性识别,验证混合预测模型的可行性；对样本序列进行噪声平滑等处理,通过构建混合预测模型对数据处理后的BDI序列进行单步和多步预测,在单步预测中分别采用传统的单独参数优化方法与基于遗传算法的参数联合优化进行仿真实验,采用遗传算法进行参数的优化选取,提高了SVM混合模型的预测能力。通过与ARIMA模型和神经网络模型进行比较,预测结果分析表明,SVM混合模型子啊BDI序列的单步和多步预测中具有较高的预测精度,能够更好地把握运价指数的变化趋势。
[Abstract]:As a barometer of dry bulk shipping market, the index of dry bulk freight rate reflects the level of freight rate in dry bulk transportation market. Due to the influence of many factors, the dry bulk freight rate index has been fluctuating sharply in recent years, and the trend is difficult to figure out, showing complex nonlinear characteristics, so it is difficult for the traditional forecasting methods to obtain good prediction results. This has also given dry bulk shipping market operators decision-making difficulties. The price index of dry bulk goods fluctuates sharply, which contains the evolution information of international dry bulk shipping market for a long time. On the basis of deep analysis of the inherent law and external influence of the fluctuation of dry bulk freight rate index, a hybrid forecasting model combining chaotic time series analysis and (Support Vector Machine, SVM) regression principle of support vector machine is proposed in this paper. The dry bulk freight rate index (Baltic Dry Index, BDI) is effectively forecasted. This paper first analyzes the supply and demand market of international dry bulk shipping and reveals the internal reasons of the fluctuation of freight rate in dry bulk shipping market. Secondly, the cause of formation and route composition of dry bulk freight rate index are expounded, and the influencing factors and fluctuation of freight rate index are qualitatively analyzed, which lays a foundation for choosing appropriate forecasting methods. In view of the nonlinear characteristics of dry bulk freight rate index, this paper presents a phase space reconstruction model combined with chaotic time series analysis and a hybrid prediction model based on support vector machine (SVM). The prediction principle and modeling idea of the hybrid model are discussed and expounded. Then, based on the systematic analysis of the selection of the key parameters of the hybrid prediction model, the mathematical model of the joint parameter optimization problem is established, and the genetic algorithm is used to solve the optimization problem. Finally, the BDl monthly mean is selected for empirical analysis to identify chaos in the BDI sample sequence to verify the feasibility of the hybrid prediction model. The sample sequence is processed by noise smoothing, and the BDI sequence after data processing is predicted by constructing a mixed prediction model. The traditional single parameter optimization method and the parameter optimization based on genetic algorithm are used to simulate the single step prediction, and the genetic algorithm is used to optimize and select the parameters. The prediction ability of SVM hybrid model is improved. Compared with the ARIMA model and the neural network model, the prediction results show that the single-step and multi-step prediction of the BDI sequence with the SVM mixed model has higher prediction accuracy and can better grasp the variation trend of the freight rate index.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：U695.27;F551;F224

【参考文献】