基于ARIMA及LSTM模型的股票分析

Stock Analysis Based on ARIMA and LSTM Models

对金融时间序列数据的研究一直广受关注,特别是股票的价格研究。文章以上证指数的开盘价为研究对象,运用ARIMA模型、ARIMA-LSTM模型以及ARIMA和ARIMA-LSTM组合模型对股票开盘价进行10天、50天、116天预测,计算每个模型的拟合优度R2,平均绝对误差MAE和均方根误差RMSE。通过比较三个模型的三个统计指标,最后得到在10天预测值时,ARIMA模型预测较好,当预测时间加长时ARIMA-LSTM模型以及ARIMA和ARIMA-LSTM组合模型表现比ARIMA模型好。

The research on financial time series data has always received widespread attention,especially in the research on stock prices.Taking the opening price of the Shanghai Securities Composite Index as the research object,this paper uses ARIMA model,ARIMA-LSTM model,and ARIMA and ARIMA-LSTM combination model to predict the opening price for 10 days,50 days and 116 days,and calculates the R2,MAE and RMSE for each model.By comparing the three statistical indicators of the three models,it is found that the ARIMA model predicts better at 10 days.When the prediction time is extended,the ARIMALSTM model and the ARIMA and ARIMA-LSTM combination model performs better than the ARIMA model.