Hybrid Long Short-Term Memory prediction model improved by particle swarm optimization with sine and cosine factors

Abstract

The Long Short-Term Memory network of deep learning neural network is widely used to predict stock price in financial field. In order to optimize the accuracy of stock price prediction by LSTM network, this paper firstly uses principal component analysis method to extract various influencing indexes of stock. Then, use Circle mapping method to select the initial value more evenly, use sine and cosine factors to improve factors of particle swarm optimization algorithm, so as to find the optimal parameters of LSTM model more effectively. Finally, the optimization results of IPSO algorithm are substituted into the LSTM model for the regression prediction with the principal components. Through empirical analysis and comparative test, the results show that the improved particle swarm optimization algorithm proposed in this paper has better optimization effect, is not prone to local optimal problems, and the prediction model based on this method has higher prediction accuracy.