| 摘要: |
| 本文构造了隐马尔可夫模型、CNN、LSTM、支持向量机的组合算法，旨在寻求股票收盘价精准预测的算法。在处理缺失值和异常值的过程中，本文使用三次样条插值法填充了缺失值，使用DBSCAN聚类的方法删除了异常值。考虑到不同的因素指标对下一交易日收盘价的影响程度不同，本文采用灰色关联判别分析其关联度，剔除了关联度小于0.9的指标，避免了数据冗余，提高了运算效率。本文使用中信证券和上证指数的数据实证研究后发现，单独预测模型并不能很好地预测收盘价的涨跌。为此，本文使用主成分分析来确定四种方法的权值，最终得出组合预测股票涨跌的比例达到95.65%和94.26%。 |
| 关键词: 主成分分析；HMM模型；CNN模型；LSTM模型；SVM模型 |
| DOI： |
| 分类号:C81 |
| 基金项目: |
|
| Forecasting Method of Stock Closing Price Combination Based on Principal Component Analysis |
|
WEI Jian，ZHAO Hong-tao，JIA He-ping
|
|
School of mathematics and physics，North China Electric Power University School of economics and management，North China Electric Power University
|
| Abstract: |
| The purpose of this paper is to find out the algorithm for accurate prediction of stock closing price，and construct a combination algorithm of hidden Markov model (HMM)，convolution neural network (CNN)，long short memory neural network (LSTM) and support vector machine (SVM).Due to the existence of missing values and outliers，in the process of filling missing values，cubic spline interpolation method is used to fill in missing values，and DBSCAN clustering method is used Method.Considering that different factors and indexes have different influence on the closing price of the next trading day，the grey correlation discriminant analysis is used to eliminate the indexes less than 0.9 to avoid data redundancy and further improve the operation efficiency.Through the empirical analysis of CITIC Securities and Shanghai stock index，it is found that the single prediction model can not predict the rise and fall of the closing price very well.Therefore，principal component analysis (PCA) is introduced to allocate the weight of four methods.Finally，through the analysis，the proportion of the rise and fall predicted by the combination model is 95.65% and 94.26%. |
| Key words: Principal Component Analysis；HMM model；CNN Model；LSTM Model；SVM Model |
