Predicting stock market price fluctuations

Abstract

The aim of this project was to create a software artefact that will, to a certain degree of accuracy, be able to predict fluctuations in the stock market prices. The particular signal that was analyzed is the S&P 500 stock market index. The first approach was that of using traditional time series analysis techniques, in the form of technical indicators, to monitor the market index. Each indicator gives their own indication of whether the stock market prices are going to rise or fall, and issues their own signal on whether it would be a good time to open or close a buy or sell position. The second approach used machine learning in the form of an Artificial Neural Network. A large amount of training data was fed to the network in an attempt for it to learn and detect patterns, with the aim of it to be able to predict stock market price movements without being given any rules or indications. The approach and accuracy of this system was then compared to other systems trying to perform a similar task using machine learning. Finally, the two approaches were combined into a hybrid trading system, and a simulation run to determine whether the combinatory approach was able to outperform either of the two trading systems.