基于GA-LSTM组合模型的股票价格预测

Stock Price Prediction Based on GA and LSTM Combination Model

随着股票市场的非线性复杂性愈加明显以及市场的波动变幻万千,传统的股票预测方法如时间序列、K线图、广义线性模型等方法已经不再适用于股票预测。为了解决传统股票预测方法中部分信息遗忘、预测精度不高等问题,提出了一种基于GA-LSTM组合模型的预测模型。采用2018年1月2日到2020年9月21日的663个交易日数据,通过GA全局寻优模型寻找LSTM预测模型的最优参数,再对沪深300股票收盘价进行预测,最后通过对比LSTM单一模型和GA-LSTM组合模型评判指标R;与MSE大小,判断模型的预测效果和预测精度。模型输出的各项指标表明,GA-LSTM的R;上升了0.0182左右,均方误差下降了0.0399左右,可以更好的用于股票预测研究。

As the nonlinear complexity of the stock market becomes more and more obvious and the market fluctuates, the traditional stock forecasting methods, such as time series, K-line graph, generalized linear model and so on, are no longer suitable for stock forecasting. In order to solve the problems of partial information forgetting and low prediction accuracy in traditional stock prediction methods, a prediction model based on ga-lstm combination model is proposed. Based on the first mock exam data from January 2, 2018 to September 21, 2020, the GA global optimization model is used to find the optimal parameters of LSTM prediction model. Then the closing price of Shanghai and Shenzhen 300 stock markets is forecasted. Finally, the prediction effect and prediction accuracy of the LSTM model are determined by comparing the LSTM single and GA-LSTM portfolio models. The results show that the R2 of ga-lstm increases by about 0.0182 and the mean square error decreases by about 0.0399, which can be better used in stock forecasting.