Application of machine learning in the forecast of stock index

• Main Author: Sugianto, Jason Jonathan
• Other Authors: Yeo Chai Kiat
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence; Engineering::Computer science and engineering::Computing methodologies::Document and text processing
• Online Access: https://hdl.handle.net/10356/156456
• Institution: Nanyang Technological University

Stock index prediction has become a hot topic and attracted a lot of researchers around the world. The stock price data constitutes of time series data which is very complex to predict due to its dynamic nature and environment. Many methods have been introduced to solve this problem, ranging from classical regression method, e.g., linear regression, etc. to newly designed machine learning method. For the last decade, the development in machine learning has inspired new opportunities and models that can be implemented in stock movement prediction domain. This has led to even larger amount of attention and research focusing specifically on this topic. This Final Year Project (FYP) focuses on predicting New York Stock Exchange Composite (NYA) index. The project is using both technical features and content features, which will be placed as an input to the model to predict the direction and value of opening and closing price’s movement of NYA index. The technical features were collected from yahoo finance website. Meanwhile, the content features were collected from various accounts in one of the most popular social media, Twitter. Content features were pre-processed to suit the model, and then converted into sentiment scores before being fed to the model. The focus of this project is the use of Long-Short Term memory (LSTM) neural network with different sets of parameters. The result will be evaluated against traditional statistical models, such as the Autoregressive Integrated Moving Average (ARIMA) model and the Vector Autoregression (VAR). The result concludes that the use of sentiment analysis has helped to improve the accuracy and performance of the model for certain set of parameters. However, with different set of parameters, the result varies slightly