支持向量机回归集成股市预测模型  

A Model for Stock Market Forecasting on Ensemble of Support Vector Regression

作　　者：汪灵枝[1,2] 赵秋梅[2] 韦增欣[2] 

机构地区：[1]柳州师范高等专科学校数学与计算科学系,广西柳州545004 [2]广西大学数学与信息科学学院,广西南宁530004

出　　处：《广西工学院学报》2010年第3期28-32,共5页Journal of Guangxi University of Technology

基　　金：广西教育厅面上项目(200707MS061)资助

摘　　要：利用基于主成分分析的支持向量机回归集成技术,提高集成个体差异度,生成一组优良的神经网络集成个体,将股票指数函数拟合成高维核空间的线性回归函数,求出一个满意的全局最优解,提高股指预测精确度,继而建立一个新型股市预测模型.试验表明,该模型能有效提高神经网络集成系统的泛化能力,预测精度高,稳定性好.In this paper,ensemble of support vector regression is implemented which is based on Principal Component Analysis Method.Then it can improve the diversity factors of the neural network ensemble individuals and generate a set of excellent ensemble individuals.Using this technology the stock index function is fitted into the linear regression function in high dimension kernel space.Meanwhile a satisfactory global optimal solution can be gotten to improve prediction accuracy.Thus a novel model for stock market forecasting is made.Tests show that the model has advantages of high accuracy,good stability and can effectively improve the generalization ability of the neural network ensemble system.

关 键 词：神经网络集成 支持向量机 主成分分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]