Font Size: a A A

The Improvement On Parameter Estimation And The Application Of Hidden Markov Models In The Financial Market

Posted on:2012-04-02Degree:MasterType:Thesis
Country:ChinaCandidate:Y ShangFull Text:PDF
GTID:2189330338984278Subject:Probability theory and mathematical statistics
Abstract/Summary:PDF Full Text Request
This paper introduces the basic construct of the Hidden Markov Model. Thismodel has been applied to many fields such as speech recognition, gene relative anal-ysis and gene recognition etc. There are three problems needed to be solved by HMM,which are training, recognizing and decoding. The theory of HMM consists of theanswers to those problems.The main work of this paper:1.Insprired by the successful application of mode recognition tools in the stockmarket, we apply the HMM model in the stock market prediction and use the historicaldata to test the model. The empirical result shows the applicability of the HMM model,but the accuracy is comparatively low.2. The historical data sequences in domestic stock market, such as daily returnsand transaction turnovers show strong serial correlation, the traditional HMM modelcan not be used to characterize the observed data sequences very well. So we introducea new modeling method in which time correlation is considered, and the model iscalled time correlated HMM, TCHMM. The empirical result shows the predictionaccuracy is improved by using the new model.3.Though the prediction accuracy is improved by the use of TCHMM model, theconvergence speed is slower in the training process caused by the increased numberof parameters in TCHMM model. The genetic algorithm is applied to the parametertraining. The simulation test and empirical results show that the speed and stability ofconvergence are improved by this algorithms.
Keywords/Search Tags:the hidden Markov model, time correlation, Baum-Welch algorithm, genetic algorithm, stock market prediction
PDF Full Text Request
Related items