Research Information System
Thesis Type: Postgraduate
Institution Of The Thesis: Middle East Technical University, Turkey
Approval Date: 2017
Thesis Language: English
Student: Mehmet Can Savaş
Co-Supervisor: YELİZ YOLCU OKUR, BÜLENT KARASÖZEN
Abstract:Algorithmic trading schemes are growing of importance in modern financial world. Each year, increasing proportion of the total trading volume is handled by algorithmic trading systems and they have become a fundamental element of modern day trading. We demonstrate the application of an algorithmic trading strategy using dynamic mode decomposition and genetic algorithm. The dynamic mode decomposition is a data analysis tool which is capable of characterizing the dynamical systems in an equation free manner by decomposing the system into low-rank structures, dynamic modes, whose temporal evolution is known. The method enables financial market prediction using dynamic modes. In order to improve the prediction success of the method, we use a complementary technical analysis tool which is optimized with genetic algorithm. We are able to build algorithmic trading strategies using dynamic mode decomposition and test them in Turkish stock market. We conclude that dynamic mode decomposition is a capable method to analyze stock markets.