基于符号序列方法的股价指数预测研究.PDFVIP

中国科技论文在线 2007 年 　7 月 　数理统计与管理 Ju l, 2007 　　　　　 App lication of Statistic s and M anagem ent 　　　　　 第 26 卷 　第 4 期 Vol26　No4 　　文章编号 : 1002 - 1566 (2007) 03 - 0602 - 08 基于符号序列方法的股价指数预测研究 苑 　莹　庄新田 (东北大学工商管理学院 ,辽宁 沈阳 110004) 摘要 :用两种符号序列方法分别对上证指数进行实证统计分析 。一种是以日收盘价计算的指数价 差的符号序列 ,另一种是以每 日 5 分钟高频数据计算的多重分形谱参数符号序列 。统计结果表 明 ,指数的涨落不是完全随机的 ,两种方法都能以一定的条件概率来预测指数的涨落 ;进一步地 , 在引入股价指数大涨落的阈值和条件平均增益后 ,发现大幅涨落时 ,条件与指数变化的关联性比 小的涨落要强得多 ,而且将两种符号序列方法结合可以更好地预测指数的大涨落 。 关键词 :符号序列 ;多重分形谱 ;股价指数 ;预测 中图分类号 : F830. 9　　　　　　　　　　　　　文献标识码 : A The Foreca stin g Research of Stock Pr ice In dex by the Sym bol Sequen ce YUAN Ying, ZHUAN G X intian ( School of Bu siness A dm in istration, Northea stern U n iversity, Shenyang City, 110004 , Ch ina) A b stract:U sing two k ind s of sym bo l sequence as the condition, th is p ap er p resen ts an emp irical re search on the data of Shanghai Stock Price Index. One is the sym bol sequence of difference of daily clo sing quotation indexe s; the other is the sym bo l sequence of m u ltifractal sp ectrum p aram eter of the h igh frequency data ticked every 5 m in itues. The re su lt show s that, the fluctuating of index is not random totally, and both of these two conditional p robab ilitie s have the foreca sting cap acity. In addition, it is found that the relation between the condition of different sym bol sequence s and change s of index in larger fluctuating of stock p rice index is stronger than the relation in sm aller fluctuating of stock p rice index when thre shold values are introduced. And the comb in ing of the se two m