Deep Learning Analysis of Australian Stock Market Price Prediction for Intelligent Service Oriented Architecture

Abstract

Stock exchanges are economic entities facilitating various trading assets like monetary values, activities, valuable metals, etc., among stockbroker participants. Prediction of Stock market rates and observing the behaviour of daily closing rates is a crucial task for many businesses and investment authorities. This acts as a precaution to know the suitable period for stakeholders to invest. Deep Learning, in this regard, is considered to perform forecasting tasks efficiently with better accuracy. For this purpose, our study performs forecasting of Australian Stock Market daily closing rates based on Deep Learning approaches of LSTM and GRU from January 4 2000, to January 17 2017. This work predicts the closing rates for the next 216 days. A comparative analysis of prediction accuracy between Deep Learning methods like Long Short-Term Memory (LSTM) along with Gated Recurrent Unit (GRU) is performed. Results reveal that the deep learning model LSTM performs better than the other approach based on the results obtained. Performance of the models is measured using metrics such as RMSE and R2 scores, where LSTM achieved a comparatively less RMSE value of 0.072 and the largest R2 score of 0.855. © 2022, ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering.