基于ARIMA与LSTM对股票价格的预测研究——以中信证券(600030)为例

Research on Stock Price Prediction Based on ARIMA and LSTM Neural Network—Taking Citic Securities (600030) as an Example

摘要: 随着我国经济的持续增长，股票市场逐渐成为整个金融业，尤其是证券业不可缺少的一部分。其中股票的价格更是引起众多投资者的关心。基于此，本文利用python在tushare上爬取中信证券(600030) 2022年3月24日到2023年3月24日一年的交易数据，并选取时间序列ARIMA模型、长短期记忆神经网络(LSTM)模型对中信证券(600030)的股票开盘价格进行训练与预测研究，实证表明ARIMA模型三天预测价格与真实价格最大误差为0.0437855，真实值与预测值非常接近；长短期记忆神经网络(LSTM)模型损失函数MES为0.33125543，并且预测值除了某几个极值点外正确率较高。这表明ARIMA模型短期预测效果较好，LSTM神经网络模型更能拟合所有的股票价格，即说明无论是ARIMA模型还是LSTM模型都能够为股票投资者提供帮助。

Abstract: With the continuous growth of China’s economy, the stock market has gradually become an indispensable part of the whole financial industry, especially the securities industry. Among them, the stock price is causing the concern of many investors. Based on this, this paper uses python to climb in tushare the trading data of Citic Securities (600030) from March 24,2022 to March 24,2023, The time series ARIMA model and long and short-term memory neural network (LSTM) model were selected to train and predict the stock opening price of CITIC Securities (600030), Empirically show that the maximum error of the three-day forecast price and the true price is 0.0437855, The true value is very close to the predicted value. Long- and short-term memory neural network (LSTM) model loss function MES is 0.33125543, and the predicted value is higher except for a few extreme points. This shows that the ARIMA model has better short-term prediction effect, and the LSTM neural network model can fit all stock prices, that both ARIMA model and LSTM model can provide help for stock investors.