基于隐马尔可夫链的上证股指建模 A Hidden Markov Chain Modeling of Shanghai Stock Index

本文引入了连续观测概率分布下的一类隐马尔可夫链模型。通过探讨该类模型的时间序列特征以及相应的模式辨识理论,提出一类新的指数预测方法。作为实证应用,我们选择上证指数2002年12月17日至2011年3月18日共2000个交易日的数据样本进行模型拟合和预测。结论显示,三维隐马尔可夫链模型能较好的拟合采样区间内的上证指数,而调整10日加权的预测方法则给出了关于样本外股指价格较为精确的预测。This paper develops a model of financial forecasting using hidden Markov chains. Empirical ana- lysis was carried out with respect to Shanghai Stock Exchange Index (SSEI) using daily data for the period from 2002.12.17-2011.3.18 (2000 samples). A three dimensional hidden Markov chain is identified to fit the data in the sampling period. An altered 10-day weighted average method was proposed, and was found to be useful for out-of-sample forecasting.