Application of machine learning in the forecast of stock index

• Main Author: Lim, Leonel Wei Jun
• Other Authors: Anwitaman Datta
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/165917
• Institution: Nanyang Technological University

The topic of stock market forecasting is currently quite popular and has drawn the attention of several experts. The stock market is renowned for its volatility, dynamism, and nonlinearity, which makes precise prediction of stock prices an exceedingly difficult task due to a plethora of macro and micro factors, including political developments, global economic conditions, unforeseeable events, a company's financial performance, and more. Numerous approaches have been proposed to address this issue, ranging from traditional regression techniques such as linear regression to recently developed machine learning techniques. Over the years, advances in machine learning have opened up new possibilities and models that can be applied to the prediction of stock market movement. This has caused this topic to receive even more attention and be the subject of further research. As part of this focus, the primary objective of this Final Year Project (FYP) is to predict the next trading day’s close price of the Standard and Poor's 500 (S&P500) index. To achieve this goal, we fed the previous 60 trading days’ open, high, low, and close prices extracted from Yahoo Finance into different models as inputs. This FYP investigates the implementation of both conventional machine learning techniques and deep learning techniques. The conventional machine learning techniques incorporated in this study are random forest regression and support vector regression, while the deep learning techniques employed are standard Artificial Neural Networks (ANN) and Long-Short Term Memory (LSTM) neural networks. The efficacy of these techniques is evaluated against each other and the final result is support vector regression (linear) is the best-performing model.