The performance analysis of stock predication based on recurrent neural network

Abstract

The stock exchange is unpredictable, and the stock price seems unpredictable. However, with the continuous development of the deep learning model's ability to deal with massive data, forecasting stock prices has become feasible and has reference value for investors. Many factors affect the stock price, and it is a great challenge to define these factors' influence on the price clearly. This paper selects multi-features stock price data sets of different companies. Because of the superiority of recurrent neural networks in dealing with time series problems, this paper compares and analyzes the experimental results of four models, namely Long Short Term Memory, Bi-directional Long Short-Term Memory, Gate Recurrent Unit, and Bi-directional Gate Recurrent Unit, and concludes that the BiLSTM model is the most outstanding one. At the same time, the prediction accuracy under different feature numbers is compared. The experimental results show that the stock price forecasting model with multi-features shows good performance, but the noise brought by it can't be ignored.