Value Investing and Machine Learning
Details
- Authors: Serhad Erdogan
- Title: Value Investing and Machine Learning
- Supervisior: Prof. Dr. Jörg Osterrieder
- Degree: Bachelor of Science
- University: University of Applied Sciences Bern, BSc Business Information Systems
- Year: 2023
- Status: Working Paper
Summary
The thesis discusses the applicability of Benjamin Graham's value investing strategy in today's stock market, considering the abundance of strategies and mentors available. It highlights the role of Graham and his successful student Warren Buffet in popularizing value investing. The thesis aims to implement value investing as an algorithm and test it using historical data. It also explores the development of a new trading strategy using artificial intelligence. The results show that the AI-generated strategy yields more profit than Graham's strategy, while still incorporating elements of value investing. The thesis concludes with recommendations to analyze the machine's decision criteria more precisely and to test the newly evaluated strategy using real-time data.
Abstract
Many people try their luck or skills in the stock market. However, the multitude of strategies and mentors leaves the question open as to which approach leads to success.
One personality who acted as a mentor and defined a strategy was Benjamin Graham. In addition to his famous strategy, value investing, he also excelled thanks to his highly successful student, Warren Buffet. Both were instrumental in popularizing value investing. However, much time has passed since the publication of the strategy. Today, huge amounts of information and data are processed almost in real-time by powerful computers. The situation has therefore changed significantly. This thesis examines whether Benjamin Graham's value investing is still applicable today.
Within the framework of this bachelor's thesis at the Bern University of Applied Sciences in the Bachelor of Science in Business Informatics program, value investing is implemented as an algorithm and tested with historical fundamental and price data. Regardless of the results, an additional trading strategy will be identified and tested with the available data.
However, the development of this new strategy is not carried out by human hands but by artificial intelligence. By providing a multitude of fundamental data to this intelligence, it is determined that it can predict a defined target value with an accuracy of 99% and capture the corresponding strategy conditions.
Testing this new strategy using historical data shows that it yields significantly more profit than Graham's strategy. However, when looking at this new strategy from a professional perspective, it quickly becomes apparent that the elements of value investing are also applied to a greater extent here.
The bachelor's thesis concludes with two recommendations: firstly, the currently rather complex decision criteria of the machine should be analyzed more precisely and a universally applicable or easily formulated rule should be extracted. Secondly, the newly evaluated strategy should be tested not only with historical data but also with real-time data (with or without real capital).