ViewFulltext-安徽建筑大学学报.doc

社会媒体环境下基于EMD-DSVR的股票市场预测方法研究 梁坤1,张理政2 （1.合肥工业大学 管理学院，合肥 230009，2.安徽建筑大学 管理学院，合肥 230022） 摘 要 现有的利用社会媒体预测股票市场的研究未能考虑股指时间序列所具有的多尺度特征。为了解决这一问题，运用EMD分解法、混沌分析理论和支持向量回归机，提出一种EMD-DSVR股票市场预测方法。首先分析股指时间序列多尺度与社会媒体变量序列多尺度间的内在联系，运用EMD分解法将社会媒体变量序列分解成不同尺度的基本模态分量；然后运用混沌分析理论和支持向量回归机对各模态分量进行建模和预测；最后利用社会媒体变量序列的各模态分量对股票市场进行预测。运用所提出的EMD-DSVR模型，对上证指数和深成指数的日收盘值进行预测，实验结果表明，所提出的方法能有效提高对股指时间序列的预测精度。 关键词 经验模态分解；股票收益；混沌理论；支持向量回归 中图分类号：TP391 文献标识码：A Analyzing Market Performance via Social Media: A EMD-DSVR Prediction Method LIANG Kun,ZHANG Lizheng (1.School of Management, Hefei University of Technology, Hefei，230009, China, 2.School of Management，AnHui Jianzhu University，HeFei，230022，China) Abstract: The existing relevant research of social media-based market performance analysis fail to consider the multi-scale of stock time series. To solve this problem, following empirical mode decomposition(EMD), chaos theory and support vector regression, a EMD-DSVR method is presented to predict stock market. First, the intrinsic links between stock time series multi-scale and social media time series multi-scale is analyzed; using EMD method, the social media time series is decomposed into many intrinsic modal function(IMF) which can significantly represent potential information of original time serial. Then, by using chaos theory and support vector regression, each IMF is modeled and predicted. Finally, market performance is predicted by using the IMF of social media time series. In order to verify the effectiveness of EMD-DSVR model, the close value of Shanghai Composite Index and Shenzhen component index are predicted by using this model. The results show that our approach can effectively improve the prediction accuracy of stock time series. Keywords empirical mode decomposition; stock time series; chaos theory; support vector regression 0 引言 股票市场是金融市场的重要组成部分，也是国家经济运行状况的晴雨表。合理分析并预测股票市场不仅能够引导股民进行正确投资，还可以为政府提供股票市场宏观调控的理论依据，进而保障经济平稳健康发展。 传统的股票市场分析与预测主要有基本面法和技术分析法[1]。随着论坛、博客和内容社区等社会媒体的快速发展与广泛应用