【摘要】 针对金融时间序列一般具有非线性、非平稳性、高信噪比和有限样本等特点,将模糊支持向量回归机引入到金融时间序列预测中.设计一种综合模糊隶属度函数,充分考虑到三点:第一噪音会导致错误的回归;第二越靠近预测点的样本对回归的影响越大;第三,离回归线越远的样本,对回归的贡献越大.综合隶属度函数,尽量剔除噪音并给离回归线远的和靠近预测点的样本较大的权值.将采用综合隶属度函数的模糊支持向量回归机应用于羊绒价格序列中,仿真结果表明,本文的基于综合隶属度函数的模糊支持向量回归机在预测精度上有所提高.更多还原

【Abstract】 Fuzzy support vector regression machine is introduced to financial time series forecasting for financial time series have the features such as nonlinear,non-stationary,high signal-to-noise ratio and so on. This paper introduces an integrated membership function which considers three points. The first,noise leads to error regression; second,effect on the regression is greater when the sample is closer to forecasting points; third,the sample far from the regression line,the greater contribution to the regression. Comprehensive algorithm of membership function tries to eliminate the noise points and gives the bigger weight to the sample which is far from the regression line or closer to the forecasting samples. Fuzzy support vector regression which uses comprehensive membership function is applied in cashmere price series,the simulation results showthat the forecasting result based on fuzzy support vector regression based on comprehensive membership function has higher prediction accuracy.更多还原