基于PCA、ICEEMDAN和LSTM的股票价格预测混合框架  

A Hybrid Framework for Stock Price Prediction Based on PCA, ICEEMDAN, and LSTM

作　　者：刘玉昆 

机构地区：[1]长沙理工大学数学与统计学院,湖南 长沙

出　　处：《应用数学进展》2023年第12期5175-5185,共11页Advances in Applied Mathematics

摘　　要：投资者在市场上买卖股票的目的是为了获得最大的回报。然而,股票价格表现出非线性和非平稳性,难以准确预测。为了解决这个问题,结合主成分分析(PCA),完全自适应噪声集合经验模态分解(ICEEMDAN)和长短期记忆网络(LSTM),制定了一个混合预测模型,称为PCA-ICEEMDAN-LSTM,以预测中国股票指数收盘价。在这项研究中,选取8个股市中常用的技术指标作为原始特征,利用PCA筛选出最符合预期的几个技术指标作为LSTM的输入特征,ICEEMDAN分解得到的分量作为目标变量。对2018~2022年中国股票市场价格的实验进行了研究,并使用各种统计指标作为评估标准。实验得到的结果表明,该框架产生了最好的性能相比,基线方法预测股票市场价格。此外,采用PCA和ICEEMDAN可以帮助提高基线LSTM模型的性能。The purpose of investors buying and selling stocks in the market is to achieve the maximum return. However, stock prices exhibit non-linearity and non-stationarity, making accurate prediction chal-lenging. To address this issue, a hybrid forecasting model, named PCA-ICEEMDAN-LSTM, was de-veloped by combining Principal Component Analysis (PCA), Improved Complete Ensemble Empiri-cal Mode Decomposition with Adaptive Noise (ICEEMDAN), and Long Short-Term Memory networks (LSTM) to predict the closing prices of Chinese stock indices. In this study, eight commonly used technical indicators in the stock market were selected as the original features. PCA was utilized to filter out the most relevant technical indicators as input features for LSTM, and the components ob-tained from ICEEMDAN decomposition were used as target variables. Experiments were conducted on the Chinese stock market prices from 2018 to 2022, and various statistical indicators were used as evaluation criteria. The results obtained indicate that this framework produces the best perfor-mance compared to baseline methods in predicting stock market prices. Furthermore, the use of PCA and ICEEMDAN helps to enhance the performance of the original LSTM model.

关 键 词：主成分分析 经验模态分解 长短期记忆网络 

分 类 号：F83[经济管理—金融学]