Neural Network Prediction Model Based On Attention Mechanism

The accurate prediction of stock prices is important.On the one hand,it can help decision makers to judge the current economic situation,on the other hand it can help investors to obtain income.However it is very difficult to make the precise prediction of stock prices.Stock prices are non-stationary,non-linear and random walk time series,which is difficult to fit with traditional time series models.With the rapid development of neural networks,the powerful learning ability of neural networks has made them successful in many fields.Many scholars have developed various neural network models for stock market prediction,including multi-layer perceptron models,convolutional neural Networks models and recurrent neural network models that specifically predict time series,etc.But these models have problems in feature extraction,low prediction accuracy and weak explanatory power.In order to solve these problems,this paper introduces the attention mechanism into the recurrent neural network model.The attention mechanism can give different weights to the information of each time dimension and distinguish the importance of different information for prediction,thereby improving the interpretability and prediction precision of the recurrent network model.In the empirical study,this paper constructs a multi-layer perceptron model,a convolutional neural network model,a recurrent neural network model,a long and short-term memory neural network model,a gated recurrent unit network model,a recurrent neural network model based on attention mechanism,a long-term and short-term memory neural network model of the attention mechanism,a gated recurrent unit neural network model based on the attention mechanism to predict the closing price based of the Shanghai Composite Index.A large number of comparative tests have been done under different parameter conditions.On the one hand,this paper explores the influence of attention mechanism on the prediction performance of the neural network model,on the other hand,this paper gives a specific optimization direction for the neural network-based stock prediction model.The empirical results show that the recursive network model based on the attention mechanism can achieve focused learning,improve learning efficiency and the accuracy and stability of prediction.