基于神经网络LSTM的Markowitz扩展模型的投资组合优化  

Optimization of Investment Portfolio Based on the Extended Markowitz Model with LSTM Neural Network

作　　者：吴楠 WU Nan(Amazon.com,Inc.,Beijing 100025,China)

机构地区：[1]亚马逊公司,北京100025

出　　处：《现代信息科技》2025年第5期159-163,共5页Modern Information Technology

摘　　要：文章对长短期记忆网络(LSTM)模型在预测股票价格方面的应用进行了研究,并探讨了如何将LSTM模型的预测结果融入Markowitz传统投资组合优化模型中。报告了LSTM模型在投资组合管理中的新现状,特别是在预期收益波动率的预测方面。通过数据集调整和训练次数的优化实验,研究对模型预测精度的提升潜力进行了调查,并发现精确度可接近90%。最后,文章基于LSTM预测数据进行了最佳投资组合构建及其收益分析。The paper conducts research on the application of the LSTM network in predicting stock prices and explores how to integrate the prediction results of the LSTM model into the traditional Markowitz portfolio optimization model.It reports the new status of the LSTM model in investment portfolio management,especially in the prediction of the volatility of expected returns.Through experiments on dataset adjustment and optimization of the number of training sessions,the research investigates the potential for improving the prediction accuracy of the model and finds that the accuracy can approach 90%.Finally,the paper conducts an analysis of the construction of the optimal investment portfolio and its returns based on the data predicted by the LSTM model.

关 键 词：人工智能 机器学习 神经网络 股票价格预测 投资组合优化 MARKOWITZ 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]