Gaussian Process Regression Based on Stochastic Segmentation of Data and Its Application to Stock Price Prediction

Abstract

Gaussian Process Regression (GPR) is a powerful model for stock price prediction in both research and practice within financial markets. However, when applying GPR to stock price prediction, the model faces overfitting and underfitting problems. Moreover, when the amount of processed data is large, computational complexity and resource consumption become significant factors limiting its practical application. To address the above challenges, this study proposes a Gaussian process regression model based on stochastic data segmentation. It aims to optimize the computational efficiency of the model and improve the prediction performance. This method not only significantly reduces the computational complexity, but also improves the prediction performance through integrated learning of sub-models. It also improves the generalization ability of the model through the integrated learning of sub-models. In addition, the introduction of the model averaging strategy effectively mitigates the overfitting and underfitting problems by weighting the uncertainty measures of the sub-models. To verify the effectiveness of the proposed method, this study first analyzes a large number of simulation experiments. The performance of the model is systematically evaluated. Secondly, by selecting the stock data of listed companies in a number of different industries as the research object, the application value and robustness of this method in the real market environment are further confirmed.