The Multifractal Analysis And Prediction Research On Stock Price

With the rapid development of science and technology, nonlinear science and its corresponding complexity problems have become a hot research topic. In the financial market at present, the data is changeable swiftly such as stock price, gold price and so on. The domestic and international scholars begin to research the regularity of the financial data in order to find some predictable methods of financial market.According to this complex data, some scholars found that the variety regulation of financial market isn't strictly random. In fact, the variety of financial data doesn't disorder but have some relativity to a certain extent. So we should look for a method that can describe and predict financial market variety regulation. Besides that, the method which can make more accurate estimate and prediction of the financial market regulation is of vital importance.Based on nonlinear theories in this paper, Fractal theory and intelligence compute methods which can analyze and compare the predictable methods of the stock price are applied in the stock market. Meanwhile, it provides some references for the financial governors and immense investors.The main researches of this paper include the following contents:(1)Fractal theory and fractal method are introduced, and the method's steady which is based on multifractal detrended fluctuation analysis of overlappe sliding window is analyzed.(2)The statistics physics computational methods of multifractal spectrum are introduced, and the fluctuation trendency of two estate stock prices are analyzed by multifractal spectrum and it's parameters.The result is that:Inχq(ε)～Inεof the stock price time sequence in the two companies have a fine linear relationship. The analysis of the changable characteristics of multifractal spectrum in stock price fluctuation has a certain predictability.(3)The prediction of stock price is researched in the paper. The paper establishes BP neural network model and RBF neural network model. Meanwhile, it analysis and compares these models. At last, a combination model is established. We can conclude that the prediction of combination model is much better than the prediction of single model.Then the relationship between multifractal spectrum parameters and stock returns is established.And then using these prediction models to forecast stock returns,and also play a very good prediction.