Evaluation Model And Algorithm Research Based On Financial Data

Since its establishment, the financial market has been a place of multilateral game, where gradually formed many kinds of analysis methods. Such as fundamental analysis method based on the national policy environment and the relationship of supply and demand; or technical analysis method simply based on data indicators.As the improving of the computer performance, quantitative transaction based on high frequency indicators became possible. As the rapid development of domestic financial market in recent years, investors’ enthusiasm is higher and higher, which results in a large amount of the transaction data and high transaction frequency, and transaction speed therefore needs to be higher and higher, and correlation between markets is becoming more and more high, and the proportion that quantitative trading accounts for is also more and more big, and the proportion of model algorithm is used to quantify transaction in the financial deals is becoming more and more big.Since the model of mean-variance established in 1952 which first introduced mathematical tools to financial research, neural networks, pattern recognition, decisive tree method are introduced to quantitative trading up to now, Quantitative trading has been developing rapidly, Studies on financial time series find out that, the guidance of the historical market to the future market behavior, by pattern recognition, Hidden Markov model has good performance in the field of multiple data recognition. This paper is aimed at study on this kind of system, which was the application of Continuous Hidden Markov model and algorithm research based on financial data. The main contents and results are as follows:First, this paper made a specific study on the development of trading algorithms at home and abroad, pointing out the disadvantages of some algorithms in the financial future market and the characteristics of the domestic market, put forward using Continuous Hidden Markov model for research on financial markets.Second, this paper made a detail analysis on the development and algorithm theory of Continuous Hidden Markov model system, mainly introduced the three processes of Continuous Hidden Markov model: identify process、decoding process and learning process, and to study the application of three types of process in the financial data. For some financial data, we use classification selection method, wavelet denoising, building a composite index according to the structure of the stock price index construction method to set the sample data and various parameters of the model system, In order to improve the performance of the model.Third, the whole model system’s processing method in the domestic market was introduced in detail, to validate the entire system from the actual transaction data. The model system studies by identifying and decoding process to research the adaptability of the model. Concluded that Continuous Hidden Markov model can identify patterns in financial data characteristics well. On this basis, we use the model of learning process in strengthening our financial characteristics and trends in the data, the results are reflected in the model is updated. Finally the updated model was forecasted the future trends of Shanghai and Shenzhen 300 stock and Futures data, and trading simulation. Through researching and comparing the simulation transaction success rate, cumulative revenue, number of transactions and so on. Concluded that the Continuous Hidden Markov model have better performance of improving the cumulative gain, reduction in the number of transactions and reduce cost shocks.Through the study of this article, put forward the overall application logic of a Continuous Hidden Markov model in the financial data, and through the analysis and comparison, proven the model have effect on feature extraction and prediction of future trends in financial data. Research results has some provide support for the theory and practice of rapid and steady development of the quantitative trading strategies and the development of algorithms.