变系数组合预测模型及其在汇率中的应用分析-variable coefficient combination forecasting model and its application analysis in exchange rate.docxVIP

论文题目z变系数组合预测模型及其在汇率中的应用研 究学科名称z应用数学研究生z王雁 指导教师z张德生教授签名z王雁签名z 了各倪摘要汇率在世界经济中起着举足轻重的作用，它是两种货币之间兑换比率的一种体现。本 文主要研究了利用函数系数自回归模型 (FAR)、自适应函数系数自回归模型(AFAR) 、考虑 观测误差的 EV 模型以及 BP 神经网络组合模型、最优加权组合模型、基于误差平方倒数 的变权组合模型、基于娟权和误差平方倒数的变权组合模型对人民币/美元汇率进行预测， 并将单个模型和组合模型的预测结果进行了比较，具体内容如下 :1.对人民币/美元汇率预处理后的序列建立了 Fi仪 模型和推广的函数系数自回归模 型，实证研究结果表明 :推广的函数系数自回归模型的拟合效果和预测精度要优于 FAR 模型的拟合效果和预测精度。2. 将 AFAR 模型和测量误差模型相结合建立了自适应变系数 EV 模型，并对人民币/ 美元汇率进行研究。实证研究结果表明:此 EV 模型的拟合效果比 AFAR 模型的拟合效果 好，其预测精度也优于 AFAR 模型。3. 对人民币/美元汇率建立了最优加权组合模型、BP 神经网络组合模型和基于误差 平方倒数的变权组合模型以及基于娟权和误差平方倒数的变权组合模型。实证分析泰明 : 组合模型的预测精度要优于单个模型;变权组合模型的预测精度优于最优组合模型;基于 娟权和误差平方倒数的变权组合预测精度要优于基于误差平方倒数的变权组合的预测精 度。关键词: 人民币/美元汇率 ;FAR 模型; AFAR 模型 ;EV 模型:变权组合AbstractTitle: THE COMBINATION OF VARIABLE COEFFICIENTS MODELS AND ITS APPLICATION IN THE EXCHANGE RATEMajor:Applied MathematicsName:Yan WANGSignature: _________ Supervisor:Prof. Desheng ZHANGSignature: _________AbstractThe exchange rate is an embodiment between two currencies exchange and plays an important role in the world economy. This dissertation mainly research functional coefficient autoregressive model (FAR), adaptive functional coefficient autoregressive model (AFAR), taking into account the observation error EV model and BP neural network combination models, the optimal weighted combination models, based on the error inverse square of variable weightcombination models, based on entropy weight and variable weight combination models aboutthe error inverse square in the RMB/USD exchange rate prediction, and compare the predicted results between a single model and combination models, the specific contents are as follows:1. We can set up functional coefficient autoregressive model and the promotion of functional coefficient autoregressive model to the RMB/USD exchange rate after preprocessing sequence. The empirical study results show that the promotion of functional coefficient autoregressive model fitting effect and prediction accuracy is better than FAR model.2. We can also combine AFAR m