基于深度学习对金融证券市场股票价格和风险问题的研究

Research on Stock Price and Risk in Financial Securities Market Based on Deep Learning

摘要: 中国金融证券市场已逐渐成熟，吸引广泛投资者参与，股票价格受内在价值和大盘指数波动影响，投资者需面对时机选择和风险价值分析的难题。本文针对中国金融证券市场的股票价格和风险预测问题，以晋城煤业作为研究对象，构建了基于深度学习的CNN-LSTM股票价格预测模型和基于信息熵与方差的风险度量模型。CNN-LSTM模型可以通过卷积神经网络提取局部空间特征，长短期记忆网络提取时间特征，并利用分位数法求局部性顶部或底部股票价格时间区间；而信息熵–方差模型综合考虑了信息熵和方差两个指标，度量股票收益的不确定性和波动性，构建了全面的风险度量模型。实证分析表明，股票价格预测模型能够较好地预测股票价格走势，并判断出局部极值的出现时间；而风险度量模型能够合理评估股票投资风险，风险值变化与股票实际波动性相符，从而为投资者投资行为提供有效支撑。

Abstract: The Chinese financial securities market has gradually matured, attracting a wide range of investors to participate. Stock prices are influenced by intrinsic value and market index fluctuations, posing challenges for investors in terms of timing and risk-value analysis. This article addresses the issue of stock price and risk prediction in the Chinese financial securities market, focusing on Jincheng Coal Industry. We have developed a CNN-LSTM stock price prediction model based on deep learning and a risk measurement model based on information entropy and variance. The CNN-LSTM model utilizes convolutional neural networks to extract local spatial features, long short-term memory networks to capture temporal features, and employs quantile regression to identify local extreme points in stock prices over time intervals. On the other hand, the information entropy-variance model comprehensively considers both information entropy and variance, measuring the uncertainty and volatility of stock returns to construct a comprehensive risk measurement model. Empirical analysis indicates that the stock price prediction model performs well in forecasting stock price trends and identifying the occurrence time of local extremes. Meanwhile, the risk measurement model effectively assesses stock investment risks, with variations in risk values aligning with the actual volatility of stocks. This provides valuable support for investors in their investment decisions.