Stock Disaster And Stock Price Fluctuation Study Based On Topological Data Analysis And Machine Learning

Because the stock market is subject to the interaction of various factors,it is a challenge to explore the development of the stock market trend.Stock data is usually characterized by high dimensions and noise,which makes stock analysis tools need to be capable of processing high dimensions and noise resistance.This paper assumes that the stock data in high dimensional space has meaningful low dimensional topology structure.And mainly with continuous homology of topology data analysis method we can effectively extract the low dimensional topology characteristics in the high-dimensional data,and it is stable to data with noise.Therefore,this paper uses the topological data analysis method to extract the stock data in low dimensional topology characteristics.Previous researches show that the machine learning method has a good effect in the analysis and prediction of multiple financial data,so this paper tries to combine the low dimensional topology information of stocks with the machine learning method.Thus,this paper puts forward the stock market situation based on topological data analysis to predict the stock market situation at the moment of stock market disaster.Then a topological data analysis method is proposed to predict the trend of stock rise and fall.Finally,a simulated trading platform based on PyQt is built.The contributions of this paper can be divided into three parts:Firstly,the topological characteristic index of stock price is put forward to distinguish stock disaster from normal time.Two testing methods are adopted in this paper.One is to take the coefficient set obtained after the wavelet transform or Fourier transform of stock sequence data as the input of the continuous homology algorithm to acquire the Persistence Diagram,and then to observe the Persistence Image.The other is to observe the Persistence Diagram corresponding to the similarity matrix between stocks.By observing the discriminant index and significance judgment,the validity of these methods in predicting stock market disaster can be explained.Secondly,methods for predicting the rise and fall of stock prices based on topological data analysis are proposed.One method is to use topological similarity to find out the period of similar stock prices,and then to predict the trend of rise and fall of stock prices.The predicted results are almost the same as the real results.The other method is to combine the low dimensional topology features obtained from topology data analysis with machine learning to predict the rise and fall of stock prices.The performance of this method is better than that of using machine learning or other methods directly.Thirdly,PyQt is exploited to build a simulated trading platform.In this paper,we propose a k-means clustering stock selection method to avoid risks by using topological data analysis,then simulate the transaction,the results show that the method in this paper has higher yield than random stock selection.Also,various machine learning methods are integrated into the interface of the trading platform,and then the optimal trading strategy is selected for trading.The platform can directly display the rate of return.Users can select the stock symbol they want to trade in this simulated trading platform,and can also set the trading strategy and other options.