基于隐马尔科夫模型(HMM)的股票价格预测分析-统计学专业论文.docxVIP

万方数据 万方数据 Abstract With the development of economy, stock market as a “barometer” has gradually received attention. Stock investment has become an important part. The prediction researches are instructive, for national macroeconomic regulation and individual investments. However, the series of stock price have characters of diversity, linkage and complexity which increase difficulties to predict it. Based on deeply analyzing key issues in the stock market and comparing the various method of stock, this article explore the feasibility of using HMM(Hidden Markov Model) to forecast and analysis the stock price. Firstly, this article introduces the theoretical framework of HMM, including concepts, principles, three classic problem and the corresponding solution algorithms. Three classic problem are evaluation, decoding and learning problems, and the corresponding algorithm are Forward-backward, Viterbi and Baum-Welch algorithm. Secondly, the article establish HMM used in stock price prediction. Specific solution: Make stock price’s open, high, low, closing price as observation sequence, use standard BIC, OEHS criterion to determine the number of hidden states; estimate the model parameters by Baum-Welch algorithm; recognize pattern of data by the Viterbi algorithm, and use maximum likelihood estimation method to predict the closing price of next day; measure forecasting accuracy of the model by the mean absolute percent error (MAPE). Thirdly, the improved HMM is optimized on the basis of traditional model. On the one hand, it removes interference noise of data and retains useful information; on the other hand, it use multi-day weighted prediction method to predict stock price. Fourthly, use MATLAB 2010a to make empirical analysis. According to the current policy, select two stocks called Agricultural Bank of China (601288) and Minsheng Bank (600016). This part use improved HMM to predict the stock price. At the same time, use the model of HMM, ARIMA, GARCH and BP neural