Predictive analysis of stock prices using gated recurrent neural network

Abstract

Stock price prediction is a popular and prevalent field of study due the large potential profits involved. However, the stock market is difficult to predict due to the many unknown and unpredictable factors affecting the market as well as random noise. Machine learning has shown great promise in many applications and in particular, recurrent neural networks (RNN) have shown promise in time series predictions. This project will focus on gated RNNs such as LSTMs and GRUs and the on the stock prices of the largest companies in the banking industry on the Singapore stock exchange. Daily trading data, technical indicators and macroeconomic variables would be mined, calculated fed to the machine learning models to predict stock prices. Different models of different types are evaluated for their suitability for stock price prediction of Singapore bank. The results show that a GRU model with auto-regression was the most successful in predicting stock price.