基于改进的分形插值与SVM的股指预测模型  

Prediction of Stock Index Based on Fractal Interpolation and SVM

作　　者：黎红 王宏勇[1] LI Hong;WANG Hongyong(School of Applied Mathematics,Nanjing University of Finance and Economics,Nanjing 210023,China)

机构地区：[1]南京财经大学应用数学学院,江苏南京210023

出　　处：《吉首大学学报（自然科学版）》2018年第3期14-19,共6页Journal of Jishou University(Natural Sciences Edition)

基　　金：教育部人文社科规划基金(12YJAZH020);南京财经大学现代服务业协同创新中心资助项目(ZWFXT14001);江苏省普通高校学术学位研究生科研创新计划项目(KYLX16_1337)

摘　　要：为了更好地分析和预测股指时间序列的短期变化趋势,提出了一种确定分形插值自由参数的新方法,由此建立了一个改进的分形插值模型,并将该模型与支持向量机模型相结合构造混合预测模型.经R/S分析可知上海证券综合指数的日收盘数据具有长程相关性,于是将混合预测模型用于分析和预测上海证券综合指数时间序列,发现混合预测模型较其他方法具有更好的拟合效果,且在短期预测方面有更高的预测精度.In order to better analyze and predict the short-term trend of stock index time series,we propose a new method to determine the free parameters of fractal interpolation,and establish an improved fractal interpolation model.This model is the combined with the support vector machine model to establish a mixed prediction model.The daily closing data of Shanghai composite index is selected as the research object which is shown to have long-range dependence thorugh R/S analysis.The time series of Shanghai composite index are analyzed and predicted by the mixed prediction model.The empirical results show that and the new mixed model proposed in this paper has good fitting performance and higher accuracy in short-term prediction.

关 键 词：分形插值 SVM模型 股指序列 预测 

分 类 号：F224.9[经济管理—国民经济] F830.9