Historical Stock Market Data Analysis Using Deep Learning Approaches

Abstract

Deep Learning evolved from Artificial Intelligence and has been used to solve problems where Machine Learning faces dead ends. The Design and Architecture of the Neural Networks, a Deep Learning paradigm is one of the major factors in deciding how successful the technologies in Deep Learning models are implemented. The architectures have evolved with varying applications and the impact it has on its output. Out of the several architectures, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) are taken, modeled for stock exchange to predict the price levels and evaluation on its performance was taken up for study.Two historical datasets containing 5021 of financial data each were taken for the analysis. Out of these 4016 data were taken to train the data and 1005 data were taken as test data.  The accuracy and performance analysis were determined by the error metrics and the computation time is sought to determine how well the models have fit the dataset while predicting in stock exchange.