Based On The Price Of Securities Of The Hidden Markov Chain Model And Empirical Analysis

This paper complete an empirical research related with HMM and stock price forecast, try to provide a new angle from the financial engineering. By discussing the feasibility of constructing the stock price forecast model by HMM and experimenting the strength of such a method by testing several stock indexes, we attempt to provide some new methods and new ideas in the field of stock price forecast through quantitative way.HMM is a probabilistic model with parametric representation, which is used to describe the statistic characteristics of the stochastic process and displayed as a double stochastic process itself. The Markov model within HMM model is an invisible chain, which indicates the states transition and described as transitional probability matrix. The general stochastic process indicates the relationship between hidden states and observations, and described as emission probability. Unlike the models as Artificial Neural Network, HMM itself has a complete theory of probability and statistics as its foundation.The number of the hidden state is the key parameter within the model, and different number will have different interpretations of the results. The 3rd chapter used the continuous observation probability distribution HMM to model the Shanghai Stock Exchange Composite Index, in order to identify the states of the series by finding the hidden states:increase, drop and even. The experiment employs the BIC/OEHS as the testing methods, and paves the way for the next study.In 4th chapter, the observation sequence is extended to a four dimensions series, including the opening price, the highest and the lowest price, in order to recognize the move pattern of the stock price time series. The main path of forecasting is to identify the similar move pattern of the target as in the historical data pool, by the means of Maximum Likelihood Estimate. In this part, the author proposed the method of d-Day Weighting Forecast; the tools as MAPE are used to analyze the outcomes from different parameters, and then d=10 is identified as an extraordinary parameter setting.In 5th chapter, we try to find the factors with the best prediction results, based on the 10-Day Weighting Forecast HMM model and Shanghai Stock Exchange 50 ETF is chose as the experiment target. Based on the combinatory analysis of basic pattern, moving average system, interest rate, money flow factors, we are able to find three factors with the best forecast results:DDX, DDY and Moving Average 10 days.