A Performance Comparison of ANN And LSTM Neural Networks in Forecasting Stock Indices

Abstract

Over the past few years, advances in the forecasting techniques has been evolved from basic linear and non-linear statistical methods to more advanced neural networks (see for eg. Convolution neural networks, Generative adversial networks etc.). Mcnelis (2005) defined forecasting as a technique that uses historical data to make informed predictions that are predictive in deciding future trends. It requires understanding the dynamic lead-lag relationship among the concerned variables, interpreting their statistical significance, and understanding which variables (give variables here) are of key importance in predicting the market moves. The key element for a sound informed financial decision in era of heightened volatility and globalization of financial markets is better forecasting.