基于LSTM模型的腾讯控股股票价格趋势预测研究

Research on The Prediction of Tencent Holdings’ Stock Price Trends Based on LSTM Model

摘要: 本文基于长短期记忆神经网络理论，构建腾讯控股股票价格趋势的预测模型，并结合数据归一化与标准化方法对输入特征进行预处理，以提升模型的预测精度。股票价格预测作为金融时间序列分析的核心问题之一，对投资者的决策优化和风险管理具有重要意义。在实证部分，本文选取2005年3月至2025年3月的腾讯控股股票历史交易数据，通过Min-Max归一化与零均值归一化消除量纲差异，利用PyTorch框架与AdamW优化器构建了单特征(收盘价)与多特征(开盘价、最高价、最低价、收盘价、成交量)输入的LSTM预测模型。实验结果表明，多特征LSTM模型的预测效果显著优于单特征模型，模型能够有效捕捉股价的时序规律与多维特征间的非线性关系。本研究验证了LSTM在金融时间序列预测中的适用性，并通过多维特征融合进一步提升了预测精度，为投资者提供了科学的量化分析工具。

Abstract: This article is based on the theory of long short-term memory neural networks to construct a prediction model for Tencent Holdings’ stock price trend, and combines data normalization and standardization methods to preprocess input features to improve the prediction accuracy of the model. Stock price prediction, as one of the core issues in financial time series analysis, is of great significance for investors’ decision optimization and risk management. In the empirical section, this article selects historical trading data of Tencent Holdings stock from March 2005 to March 2025, and uses Min Max normalization and zero mean normalization to eliminate dimensional differences. Using the PyTorch framework and AdamW optimizer, a single feature (closing price) and multi feature (opening price, highest price, lowest price, closing price, trading volume) input LSTM prediction model is constructed. The experimental results show that the predictive performance of the multi feature LSTM model is significantly better than that of the single feature model, and the model can effectively capture the temporal patterns of stock prices and the nonlinear relationships between multidimensional features. This study validates the applicability of LSTM in financial time series prediction and further improves prediction accuracy through multidimensional feature fusion, providing investors with a scientific quantitative analysis tool.