Multi-Variate Regression Analysis for Stock Market price prediction using Stacked LSTM

Abstract

Understanding a reasonably insightful trend in stock prices is significantly important in the concerned community and stakeholders to minimize risks on investment. Unfortunately, the non-linear, volatile and unpredictable nature of stock market values makes it exceptionally challenging to forecast future trends on the stock market. The availability of substantial stock data is an advantage which, however includes additional challenges in manually analyzing large volume data. Since it is not humanly possible for the traders and investors of the stock market to understand the nature of changes in stock prices, it is therefore necessary to introduce an automated system that can forecast future stock prices with an accuracy of significant level. Thankfully, machine learning-based forecasting techniques are already in effect to predict future stock prices. In this paper, we propose an integrated solution consisting of an extended Long Short Term Memory (LSTM) model in a Multivariate feature correlation approach to exploit important relationships and further facilitate the training. Experimented with Dhaka Stock Exchange data, we present an in-depth analysis of our findings, which shows the promising potential of our proposed architecture.