基于隐马尔科夫模型的市场指数量化择时研究  

Research of Market Index Quantitative Timing Based on Hidden Markov Model

作　　者：傅中杰 吴清强[1] FU Zhongjie;WU Qingqiang＂(Software School of Xiamen University,Xiamen 361005,China)

机构地区：[1]厦门大学软件学院,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2018年第3期404-412,共9页Journal of Xiamen University：Natural Science

摘　　要：量化择时是量化投资领域的重要组成,主要负责评判何时进行交易.为了验证隐马尔科夫模型(hidden Markov model,HMM)应用到量化择时的可行性,基于股票市场原始数据计算得到候选特征集,并利用HMM对各个单特征进行特征筛选,最后使用选出的特征集训练得到综合模型,预测交易日的市场状态.实验结果表明,基于HMM的交易策略比双均线策略和基于k-均值(k-means)聚类的策略都有更好的表现,且具有较强的识别市场状态、规避系统性风险以及获取超额收益的能力.Quantitative market timing constitutes an important part of quantitative investment to choose the best trading opportuni ty.To verify the feasibility of applying hidden markov model （HMM） to quantitative market timing, we creatively calculate candidate features set based on raw data, use H MM to test performance on each single feature,and train a comprehensive model using selected features to predict the market state of the next trading day. Experimental results show that HMM based strategy enjoys better stabil ity and profitability compared with strategies based on moving average or k means. Finally, HMM can skillfully identify market states ,avoid systematic risk and obtain excess return.

关 键 词：隐马尔科夫模型 市场择时 交易策略 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]