Stock Price Forecast Based on LSTM Neural Network

Abstract

LSTM (long-short term memory) Networks is one of the RNNs(recurrent neural networks). The algorithm was first published at Neural Computation by Sepp Hochreiter and Jurgen Schmidhuber. It performs better than normal RNNs in processing and predicting time series related data. At present, LSTM has achieved considerable success on many issues and has been widely used. Based on the excellent performance of LSTM Networks in time series, this article seeks to investigate whether LSTM can be applied to the stock price forecast. This paper compares the pros and cons of LSTM in time series prediction by comparing RNNs with LSTM. In this paper, the daily data of the Shanghai Composite Index and the Dow Jones Index is taken as the research object, and RNNs and LSTM are respectively used to construct the model. The criterion of the pros and cons of the model is the mean square error between predicted value and real value. This paper finally finds that LSTM can be well used in stock price forecasting.