基于ARIMA-LSTM组合模型的股价预测

Stock Price Prediction Based on ARIMA-LSTM Combination Model

本文的研究基于对浦东金桥股票1799个交易日的时间序列数据进行分析,旨在利用ARIMA-LSTM组合模型对该股票的收盘价进行精确预测。通过网格调参确定了ARIMA模型的最佳阶数,以确保模型能够有效捕捉时间序列数据中的趋势和周期性变化。利用这一优化后的ARIMA模型对浦东金桥的收盘价进行预测,从中获得预测残差。而后将残差数据输入到LSTM (长短期记忆网络)模型中。LSTM作为一种适合处理序列数据的深度学习模型,能够更好地捕捉数据中的长期依赖关系和非线性动态。通过结合ARIMA模型的残差和LSTM模型的预测能力,构建了一个ARIMA-LSTM组合模型,进一步提升了对浦东金桥收盘价未来走势的预测准确性和稳定性。This study is based on the analysis of 1799 trading days’ time series data of Pudong Jinqiao stock, aiming to use the ARIMA-LSTM combination model to accurately predict the closing price of the stock. The optimal order of the ARIMA model was determined through grid tuning to ensure that the model can effectively capture trends and periodic changes in time series data. Use this optimized ARIMA model to predict the closing price of Pudong Jinqiao and obtain the prediction residual. Then, the residual data will be inputted into the LSTM (Long Short Term Memory Network) model. LSTM, as a deep learning model suitable for processing sequential data, can better capture long-term dependencies and nonlinear dynamics in the data. By combining the residual of the ARIMA model with the predictive ability of the LSTM model, an ARIMA-LSTM combination model was constructed to further improve the accuracy and stability of predicting the future trend of the closing price of Pudong Jinqiao.