Research on stock prediction based on simulated annealing algorithm and ensemble neural learning

Abstract

By combining the relevant theoretical knowledge of simulated annealing algorithm, the parameters of the XGBoost model are optimized, and the LSTM-SA-XGBoost stock prediction model is designed. In order to verify the prediction performance of the combined model, the combined model and the single model are respectively evaluated by the prediction model performance evaluation indicators: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Accuracy rate (Accuracy) and the f1-score is analyzed and evaluated to verify that the two combined models proposed in this paper have good approximation ability and generalization ability in the prediction of stock fluctuations, and can improve the prediction performance of a single LSTM model or a single RNN model. The prediction performance of stock price rise and fall is verified, and the hybrid model based on LSTM neural network proposed in this paper has certain feasibility and stability in the prediction of stock price rise and fall.