影响伦敦金价格相关因素的实证分析

发布时间：2018-03-30 13:07

本文选题：黄金价格　切入点：小波去噪　出处：《兰州大学》2013年硕士论文

【摘要】：随着金融市场中金融产品的多元化,以及金融衍生品的多样化,人们会越来越加大在金融产品上的投资,但鉴于金融产品的特点,收益既伴随着风险,显示出预测准确的重要性。众多学者开始对黄金价格进行预测,黄金价格序列受许多因素的影响并且具有较强的随机波动性和一些尖峰,因此,用单一的时间序列模型例如R、MA、ARIMA进行预测往往不太理想,所以本文选择了小波神经网络来进行预测黄金价格。一方面利用了小波去噪使数据平滑,这样去掉干扰大的噪声,另一方面神经网络具有很强的非线性拟合能力,可映射任意复杂的非线性关系,所以在神经网络的基础上加上小波去噪便可以更好地预测黄金价格的波动。从而为投资者提供参考。主要内容如下： 1.在参阅文献和理论的思考上,分析可能影响黄金价格的元素,进而对他们用PERSON相关系数进行分析,找出了美元指数,原油,银,道琼斯指数,上征指数五大因素,并确定他们的相关系数。 2.对选取的伦敦金价格进行去噪,然后再把对应的美元指数,原油,银,道琼斯指数,上证指数的数据代入,建立去躁前和去噪后的神经网络模型,再分别确定预测值,和真实值作对比。 3.然后分别算出去噪前模型的和去噪后模型的RMSE、MAXAE、MAE、MAXAPE、 MAPE,充分用数据说明小波神经网络对于黄金价格预测拟合度好,准确度高。
[Abstract]:With the diversification of financial products in financial markets and the diversification of financial derivatives, people will increase their investment in financial products more and more, but in view of the characteristics of financial products, the returns are accompanied by risks. Many scholars began to predict the gold price. The gold price sequence is influenced by many factors and has strong random volatility and some spikes. It is not ideal to use a single time series model such as RMA Arima to predict gold price. In this paper, wavelet neural network is chosen to predict gold price. On the one hand, wavelet denoising is used to smooth the data so as to remove the noise with large interference. On the other hand, neural networks have strong nonlinear fitting ability, and can map any complex nonlinear relations. So adding wavelet denoising to the neural network can better predict the fluctuation of gold price and provide reference for investors. The main contents are as follows:. 1. On the basis of literature and theoretical considerations, we analyze the elements that may affect the price of gold, and then use the PERSON correlation coefficient to find out the five major factors: dollar index, crude oil, silver, Dow Jones index and index. And determine their correlation coefficient. 2. The selected London gold price is de-noised, then the corresponding data of the dollar index, crude oil, silver, Dow Jones index and Shanghai stock index are substituted, and the neural network models before and after de-noising are established, and the predicted values are determined respectively. Compare with the true value. 3. Then the RMSE / MAXAE / MAXAPE, MAPE of the pre-noise model and the de-noised model are calculated respectively, which shows that the wavelet neural network has good fitting and high accuracy for gold price prediction.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.9;TP18

【参考文献】