Gaussian Process Stock Price Prediction Based On Combined Kernel Function

Stock is one of the many ways to obtain investment income in modern society.Stock is one of the important ways for stock companies to get capital.Stock prediction has always been the focus and difficulty of research.In order to obtain returns in stock investment,people have always tried different prediction models to predict the stock price and its trend.A good stock price prediction model can help shareholders obtain more investment returns.Time series model is widely used because of its low time complexity and effective short-term prediction.In recent years,machine learning has been widely studied and applied.Many machine learning algorithms have been applied to financial problems,such as random forests,neural networks and Gaussian process regression.In this thesis,the data-driven method is used to form a new combination kernel function according to the characteristics of Vanke stock price,and the particle swarm optimization algorithm is used to optimize the hyper-parameters of the kernel function.Based on the new combination kernel function,the Gaussian process regression prediction model of single index and multiindex is established,and the long-term and short-term prediction of Vanke stock price is made.The main work of this paper is as follows :(1)Aiming at the problem that there is no unified theoretical support for the selection of kernel function type and the setting of initial iteration value of kernel function hyperparameters in Gaussian process regression model,the influence of single kernel function,combined kernel function and kernel function hyper-parameters on the prediction distribution is analyzed in detail by statistical software,which provides a reference for the later modeling analysis.(2)The particle swarm optimization algorithm is used to optimize the super parameters of the combined kernel function,and the Gaussian process regression prediction model of single index stock price based on the combined kernel function is established.The stock price of Vanke Real Estate is predicted in the short term.The prediction results are compared with the short-term prediction effect of the ARIMA model.It is concluded that the Gaussian process regression prediction model has high prediction accuracy and good stability.(3)For multiple indicators of stock prices,the traditional index selection methodfiltering method is used to select several representative indicators from multiple indicators,and a multi-index Gaussian process regression prediction model is established.The particle swarm optimization algorithm is used to optimize the hyperparameters of the kernel function in the Gaussian process regression model,and the stock price of Vanke Real Estate is predicted in the long term.