Machine Learning in Financial Engineering

Student: Joshua Stabler

Supervisor: Prof. Tom Downs

Category: Computer Systems Engineering Thesis Project

In the early 1990's, two major developments into the prediction of the financial markets, the first was the large supply of empirical evidence that was found to disprove the previous long-standing theory of the Efficient Market Hypothesis (EMH). The second was the introduction and acceptance of Naive Bayes method of classification as a highly competitive machine learning technique.

The new evidence into the EMH has shown that there are inefficiencies in the markets. These null hypothesis proofs show that it is possible to exploit these inefficiencies, such as the delays in the reactions of investors to the news of unexpected profit announcements, and misalignments from the fundamentals of the markets.

This thesis has been able to show some exploitation of the inefficiencies in the results. However, due to the lack of publicly released comparisons, an indication of the success of these results is difficult to judge.

Poster Presentation (PDF)

Thesis Document (PDF)