摘要
股票市场价格预测一直以来都被认为是金融时序预测领域的一项具有挑战性的工作。综合回归支持向量机SVR和自组织特征函数(SOFM)技术,并引入基于过滤的特征选择算法确定重要的输入变量,在SVR核函数的参数选择上采用粒子群优化算法(PSO)。SOFM算法将训练样本聚类,然后分别应用SVR来预测股票价格走势。最后应用上海A股的浦发银行日数据来做股票价格日预测,实验结果表明,经过改进的SOFM-SVR模型与之前的SOFM-SVR模型相比,在预测精度和训练时间上都有了较大的提高。
Stock market price prediction is regarded as a challenging task of the financial time series prediction process.In this paper we synthesize regressive SVR with self-organizing feature map（SOFM） technique,introduce filter-based feature selection algorithm to choose important input attributes,and adopt PSO to choose the best parameters of SVR kernel function.We use SOFM algorithm to train the samples clustering,and employ SVR respectively to predict the price trend of stock.In the end the daily data of Pudong Development Bank dataset of Shanghai A share market are employed to make the daily price prediction.Experiment results show that the improved SOFM-SVR model achieves a considerable melioration over the traditional SOFM-SVR model in average prediction accuracy and training time.
出处
《计算机应用与软件》
CSCD
2010年第5期172-175,178,共5页
Computer Applications and Software
关键词
组合预测
SOFM-SVR
特征选择
PSO算法
Combined prediction SOFM-SVR Feature selection Particle swarm optimization（PSO）
