摘要
结合粗糙集的相关理论,优化了GA属性约简方法,针对上证指数预测的具体问题,对遗传算法的初始种群和适应度函数进行改进,将上证指数10年间数据的58个属性构成的训练集进行属性约简,并应用参数优化后的SVM分别以属性约简前后的数据集对开盘指数进行回归预测.仿真结果表明,用该算法进行属性约简后,原始数据集中冗余属性对预测结果的影响下降,预测精度提高,建模时间也相应的减少,得到了较好的结果.
An optimized genetic algorithm for attribute conduction which was based on rough set theories was proposed. Initial population and fitness function of the genetic algorithm were improved according to situation of Shanghai Composite Index forecasting. It performs application of attribute reduction with training set which retrieve 58 attributes data from Shanghai Composite Index in recent ten years. Then, it conducts regression prediction by using parameters optimized SVM to predict stock index on the original dataset and the simplified dataset. Emulation experiment results show that it has better predict precision and time consuming performance by using simplified dataset.
出处
《福建师范大学学报(自然科学版)》
CAS
CSCD
北大核心
2011年第5期29-33,共5页
Journal of Fujian Normal University:Natural Science Edition
基金
福建省自然科学基金资助项目(2009J01273)
关键词
遗传算法
属性约简
上证指数
SVM
genetic algorithm
attribute reduction
Shanghai composite index
SVM