【摘要】 针对传统的基于隐马尔可夫模型HMM(Hidden Markov model)的股票价格序列预测方法的不足,提出一种新的基于HMM的股票价格预测的方法。采用一种CBIC(Clustering and BIC)算法自动确定HMM隐状态数,在预测过程中当预测误差大于一定阈值时,采用模型自动更新方法建立新的模型。通过对股票价格序列的转换,建立相应的HMM,进行单步值预测。单步值预测与Hassan等人的HMM fusion model方法、ARIMA方法进行了比较,实验结果表明所提出的预测算法在股票价格预测中,比现有的不更新模型的方法能得到更好的结果。更多还原

【Abstract】 In this paper,a novel Hidden Markov model(HMM)-based stock price series forecasting method is proposed for overcoming the disadvantage of traditional HMM-based stock price series forecasting method.In method,the model selection method CBIC is applied to automatically determine hidden state number of HMM,when the forecasting mean error increases to over the predefined threshold,the method of model automatic update is used to set up a new model.Corresponding HMM has been set up by transforming the stock time series,and the single-step value forecast is performed.Comparison is made on single-step value forecast with the HMM fusion model-based method by Hassan and the ARIMA method.Experiment results show that the forecasting algorithm proposed can acquire better outcomes in stock price forecast than existing models without update.更多还原