An empirical research and comprehensive analysis of stock market prediction using machine learning and deep learning techniques

Abstract

Financial markets are inherently unpredictable. They continue to change based on the performance of the company, past records, market value and are also dependent on news & timings. By carrying out trend analysis, one has the ability to prejudge stock prices. Machine Learning Techniques that are available, have the potential to forecast future stock prices. Each stock represents a different trend, so a singular machine learning model can't be applicable to other stocks. Thus, one model giving a high degree of precision can't guarantee working on another. Too many variables are involved while predicting stocks-physical factors vs. psychological, irrational and rational behaviour, etc. All of these factors combined indicate stock prices as capricious and difficult to foresee. In order to resolve this challenge, a comprehensive study with historical data on stock prices of listed firms was performed. The main premise behind this research was to illustrate how to apply machine learning algorithms such as Averaging, Linear Regression including advanced deep learning techniques such as Long-Term Short Memory and applying technical tools like the Modern Portfolio Theory and Bollinger bands.