混合LSTM-VMD在股价趋势预测中的应用  

Application of Hybrid LSTM-VMD Model in Stock Price Trend Prediction

作　　者：陈忠源 路翀 王艺涵 龙洁 雷一鸣 CHEN Zhongyuan;LU Chong;WANG Yihan;LONG Jie;LEI Yiming(College of Information Management,Xinjiang University of Finance and Economics,Urumqi,China,830012)

机构地区：[1]新疆财经大学息管理学院,乌鲁木齐830012

出　　处：《福建电脑》2024年第11期16-24,共9页Journal of Fujian Computer

基　　金：国家自然科学基金(No.62166039)资助。

摘　　要：为提高分析和应对股市波动的能力,本文构建了一个基于LSTM-VMD的股价预测模型。在混合预测框架的基础上采用自适应VMD再分解机制来自动寻找最优的分解次数,将深度学习和信号处理相互融合,通过多次分解来减少噪声,提高模型的预测精度。实验结果表明,混合LSTM-VMD模型在多个数据集上展现出更优的鲁棒性和预测准确度,特别是在捕捉时间序列数据的长期依赖特征和减少预测误差中的噪声干扰方面表现出色。To improve the ability to analyze and respond to stock market fluctuations,this paper constructs a stock price prediction model based on LSTM-VMD.On the basis of the hybrid prediction framework,an adaptive VMD decomposition mechanism is adopted to automatically find the optimal decomposition times,integrating deep learning and signal processing to reduce noise and improve the prediction accuracy of the model through multiple decompositions.The experimental results show that the hybrid LSTM-VMD model exhibits better robustness and prediction accuracy on multiple datasets,especially in capturing long-term dependency features of time series data and reducing noise interference in prediction errors.

关 键 词：混合模型 自适应VMD 噪声处理 股价预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]