基于EMD_GA_KELM的国际原油价格预测模型研究

发布时间：2018-04-13 11:38

本文选题：国际原油价格 + 核化极速神经网络　；参考：《重庆工商大学》2015年硕士论文

【摘要】：石油是保障国民经济各部门顺利运行的重要战略物资。近十年来,国际原油价格的频繁波动越来越成为制约各国经济平稳运行的不稳定因素。中国是全球最大的原油进口国,原油价格的波动对中国经济的稳定运行造成了干扰。在变化莫测的世界原油市场中,若能对油价的走向作出正确的预测,这样在面对油价的大幅波动时,可以将其造成的不利经济影响降到最低,使自身利益得到最大限度的维护。因此,密切关注国际原油市场,探索国际原油价格变动的潜在原因,对原油价格的走向进行合理的预测,对国家、企业和个人都具有重要的意义。鉴于原油具有商品、金融及政治等多种属性,本文将WTI原油价格的月度数据作为研究对象,不仅考虑油价时间序列自身的发展规律,并且通过灰色关联分析(GRA)筛选了影响原油价格的8个重要因素,提出了EMD_GA_KELM原油价格预测模型。本文的模型构建主要体现以下两个方面的工作:首先,核化极速神经网络(KELM)是借鉴支持向量机(SVM)核函数的原理对极速神经网络(ELM)的扩展,因此KELM模型同样存在惩罚系数与核参数难以选取的问题,本文运用遗传算法(GA)对KELM的惩罚因子和核参数进行优化,建立GA_KELM预测模型,通过对WTI原油月度价格的实证分析表明,GA_KELM预测的均方误差为0.05,其预测精度较没有优化的KELM模型和SVM模型分别提高了5.7%和16.75%;其次,本文将主要用于信号技术的经验模态分解方法(EMD)运用在非平稳非线性的油价序列研究中,将原油价格序列分解成若干不同频率的分量,单独对每个分量运用GA_KELM模型进行预测,将每个分量的预测结果通过相加重构的方式得到最终的原油价格预测值,实证结果表明,采用EMD_GA_KELM进行预测的效果要远远好于单独采用GA_KELM模型预测的效果,相对误差为0.041,预测精度提高了17.5%,这也说明了本文所使用的预测方法是可行的,可以作为未来原油价格预测的有效方法之一,对原油价格的预测具有较大参考意义。
[Abstract]:Petroleum is an important strategic material to ensure the smooth operation of various departments of the national economy.In recent ten years, the frequent fluctuation of international crude oil price has become an unstable factor restricting the smooth operation of economy.China is the world's largest importer of crude oil, and fluctuations in crude oil prices have interfered with the stable operation of China's economy.Therefore, it is of great significance for countries, enterprises and individuals to pay close attention to the international crude oil market, explore the potential reasons for the change of international crude oil prices, and make a reasonable prediction of the trend of crude oil prices.In view of the commodity, financial and political properties of crude oil, this paper takes the monthly data of WTI crude oil price as the research object, not only considering the development law of oil price time series itself.Eight important factors affecting crude oil price were screened by grey relational analysis (gra), and the EMD_GA_KELM crude oil price prediction model was put forward.The model construction of this paper mainly embodies the following two aspects of work: firstly, the nucleation extreme speed neural network (KELM) is an extension of the extreme speed neural network (ELM) based on the principle of support vector machine (SVM) kernel function.Therefore, the KELM model also has the problem that it is difficult to select the penalty coefficient and kernel parameter. In this paper, the penalty factor and kernel parameter of KELM are optimized by genetic algorithm (GA), and the prediction model of GA_KELM is established.The empirical analysis on the monthly price of WTI crude oil shows that the mean square error of Gackelm prediction is 0.05, and the prediction accuracy is 5.7% and 16.75% higher than that of the unoptimized KELM model and SVM model, respectively.In this paper, the empirical mode decomposition (EMD) method, which is mainly used in signal technology, is applied to the study of non-stationary and nonlinear oil price series. The price sequence of crude oil is decomposed into several components with different frequencies, and each component is predicted by GA_KELM model.The final crude oil price prediction value is obtained by adding and reconstructing the prediction results of each component. The empirical results show that the effect of EMD_GA_KELM prediction is much better than that of using GA_KELM model alone.The relative error is 0.041 and the prediction accuracy is improved by 17.50.This also shows that the prediction method used in this paper is feasible and can be used as one of the effective methods for predicting the price of crude oil in the future. It is of great reference significance for the prediction of crude oil price.
【学位授予单位】：重庆工商大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F416.22;F764.1;F224

【参考文献】