Stock prediction using machine learning

• Main Author: Ang, Eugene Joon Kit
• Other Authors: Liu Linbo
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/140161
• Institution: Nanyang Technological University

Generally, stock investors tend to implement different analysis tools on stock prediction, in order to make money from investing in stock market. However, due to the uncountable factors that influence the fluctuation of stocks such as the world economy, political issues, natural disasters, news and corporations’ development, stocks in the future can be difficult to predict[1]. Stocks in the future cannot be predicted precisely but can be predicted strategically so that the investment in stock in long term can yield to a positive gain. A combination of different stock prediction methodologies is often used to construct a decent stock investment strategy. In general, the methodologies to predict stocks can be categorized into technical analysis, sentimental analysis and fundamental analysis[2]. In this paper, technical analysis is mainly focused on using different machine learning models to compare their performances.Technical analysis of stock is a way of analyzing stock data to predict stock trend and stock price in the future. It uses the historical market data, mainly focusing on prices and volume of stocks to build a model to predict the price movement in the future. In this paper, several machine learning architectures are implemented to compare the accuracy and speed of stock prediction, supported by providing visualizable graphs and tables of root mean square error (RMSE) for each method. This paper provides the performance and conclusion of implementing Support Vector Machine (SVM) and Recurrent Neural Network (RNN), especially on Long Short-term Memory (LSTM). To better visualize the different between each model, stock of Apple and Amazon for the past 5years will be used to find the most suitable machine learning model to predict their stock in the future. This is to make sure that the huge difference between the stocks trend in the past 5years does not cause any bias on the result of this project.