基于高斯分布隐马氏模型识别市场风格构建量化策略

Gaussian HMM-Based Quant Strategies for Market Style Identification

摘要: 本文旨在研究隐马尔可夫模型在识别市场风格以及市场状态切换的能力以及探索hmm模型在行业轮动策略中的应用。通过对A股市场股票数据的细致收集与整理，使用价量因子构建与训练基于高斯分布的隐马尔可夫模型(Gaussian HMM)，该模型能有效地解码出市场状态的隐含序列。在策略构建环节，本文提出了双策略模型，包括历史状态匹配策略和未来状态观测值预测策略。经过多轮回测检验，这两种策略在识别市场风格、把握市场趋势以及优化交易决策方面均展现出显著能力。进一步构建行业轮动模型，观测整体策略收益表现。本研究通过借助高斯分布的隐马氏模型对市场风格识别的能力输出择时交易信号，制定行业轮动决策，挖掘HMM在中国A股市场中的潜在价值，为投资者提供了新的、有效的量化投资策略。

Abstract: This study aims to investigate the capabilities of Hidden Markov Models in identifying market styles, switching market states, and exploring their application in sector rotation strategies. Through meticulous collection and organization of Chinese stock market stock data, a Gaussian HMM is constructed and trained using price-volume factors, effectively decoding latent sequences of market states. In terms of strategy construction, a dual-strategy model is proposed, encompassing a historical state matching strategy and a future state observation prediction strategy. After rigorous backtesting, both strategies have demonstrated remarkable abilities in market style identification, trend capture, and trading decision optimization. Furthermore, a sector rotation model is developed to assess the overall performance of the strategy. Leveraging the market style identification capabilities of the Gaussian HMM, this study generates timely trading signals, formulates sector rotation decisions, and explores the potential value of HMMs in the Chinese A-share market, providing investors with novel and effective quantitative investment strategies.