计算机应用研究2017，Vol.34Issue(11)：3373-3378,6.DOI:10.3969/j.issn.1001-3695.2017.11.038
基于最小方差的股市拐点预测方法
Stock turning point prediction method based on minimum variance
摘要
Abstract
It is very important to predict stock turning point for assisting investment in stock market.However,the forecast of the stock turning point is an imbalanced data classification.To solve the deviation problem caused by treating the imbalance problem with SVM,this paper proposed a method to select the penalty factors.The method defined a judge which was the product of the variances of the recall and precision of all classes after executing cross validation on the training set.It selected the penalty factors corresponding to the minimum product of the variances as the optimal penalty factors of the corresponding categories,and applied the Biased-SVM model to the prediction of stock turning point.In experiments,it selected the common stock indexes as the input vectors and compared the method with other methods to solve the imbalanced problem.Experimental results demonstrate that the minimum variance method improves the recognition accuracy of the two classes including the turning points under guaranteeing the recognition accuracy of the class with more samples.It can provide help investments.关键词
股市拐点预测/不平衡分类/最小方差法/SVM/惩罚因子Key words
stock turning point prediction/unbalanced classification/minimum variance method/SVM/penalty factor分类
信息技术与安全科学引用本文复制引用
石陆魁,秦志娇,闫会强..基于最小方差的股市拐点预测方法[J].计算机应用研究,2017,34(11):3373-3378,6.基金项目
天津市应用基础与前沿技术研究计划重点项目(14JCZDJC31600) （14JCZDJC31600）
河北省自然科学基金资助项目(F2016202144) （F2016202144）
河北省高等学校科学技术研究重点资助项目(ZD2014030) （ZD2014030）
河北省科技计划资助项目(13456243) （13456243）
