GA-SVM对上证综指走势的预测研究  

A study on prediction of market tendency on the shanghai stock index based on GA-SVM method

作　　者：张伟[1] 李泓仪[2] 兰书梅[1] 张洁[1] 

机构地区：[1]吉林大学计算机科学与技术学院,吉林长春130012 [2]吉林大学经济学院,吉林长春130012

出　　处：《东北师大学报（自然科学版）》2012年第1期55-59,共5页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家科技支撑计划子课题资助项目(2006BAJ18B02-06)

摘　　要：将支持向量机和遗传算法结合,建立了一种智能数据挖掘技术(GA-SVM),并用于对上证综指市场走势进行了探索.在这个混合的数据挖掘方法中,GA用于RBF参数的设定以及特征集的选择,从而智能的找到SVM的最佳参数,减少SVM特征值的复杂度,提高了SVM算法速度.SVM用于判断未来股票市场的走势,并与统计模型、时间序列模型方法、神经网络进行了对比.实验证明,GA-SVM优于其他几种方法,这种方法对于股票上涨或下跌的预测研究是有效的.Support vector machine is an effective data mining technology for limited sample data,genetic algorithm is an excellent tool for global optimization.In this study,a hybrid data mining model which combine support vector machine with genetic algorithm（GA-SVM） is proposed to the prediction of market tendency on the shanghai stock index.In this hybrid data mining approach,GA is used to select the RBF parameters and the features,so that to find the best parameters of SVM.That can reduce model complexity of SVM and improve the speed of SVM;SVM is used to judge the future movement direction of the stock market based on the use of historical data.To validate GA-SVM method,we compared its performance with that of other methods（such as statistical method,time series method and neural network method）.The experimental results show that GA-SVM is superior to other methods,implying that the GA-SVM approach is a promising alternative to stock market tendency prediction.

关 键 词：支持向量机 遗传算法 GA-SVM 股票走势预测 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]