数字经济时代的智能投资决策——基于多模型融合的股价预测研究

Intelligent Investment Decision-Making in the Digital Economy Era—Research on Stock Price Prediction Based on Multi-Model Fusion

摘要: 随着数字经济的迅猛发展，数字化转型已经成为经济市场不可逆转的趋势，数据是关键生产要素，算法是使数据产生价值的核心生产力，股票价格能够反映不同国际贸易情形下的不同市场表现。本文选取贵州茅台的股票价格数据以及公司财务报表中的17项关键特征数据，包含了公司的盈利情况、偿债能力、运营效率和成长效率等多个方面，全面反映公司的经营状况和未来发展潜力。利用LASSO回归模型对财务报表数据进行筛选和降维，保留最具代表性的特征。构造基于LASSO和BP神经网络的符合模型，结果表明，该复合模型的均方根误差均低于单独的LASSO模型或者单独的BP神经网络模型。证明了在数字经济背景下，基于LASSO和BP的宽深神经网络模型在数据预测方面的优越性，以及将传统财务分析方法和现代机器学习技术相结合的有效性。

Abstract: With the rapid development of the digital economy, digital transformation has become an irreversible trend in the economic market. Data is a key production factor, and algorithms are the core productivity that enables data to generate value. Stock prices can reflect the different market performances under various international trade scenarios. This paper selects the stock price data of Kweichow Moutai and 17 key characteristic data from the company’s financial statements, covering multiple aspects such as the company’s profitability, debt-paying ability, operational efficiency, and growth efficiency, comprehensively reflecting the company’s operating conditions and future development potential. The LASSO regression model is used to screen and reduce the dimension of the financial statement data, retaining the most representative features. A composite model based on LASSO and BP neural network is constructed. The results show that the root mean square error of this composite model is lower than that of the LASSO model or the BP neural network model alone. This proves the superiority of the wide and deep neural network model based on LASSO and BP in data prediction under the background of the digital economy, as well as the effectiveness of combining traditional financial analysis methods with modern machine learning techniques.