This project focuses on reducing costs and risks in algorithmic trading while maintaining profitability through machine learning.
Algorithmic trading, particularly high-frequency trading, has grown significantly since the introduction of electronic trading and now accounts for a substantial portion of U.S. equity trading volume. This form of trading is gradually replacing traditional methods, emphasizing the need for transparent, interpretable models to analyze market patterns, especially for fund managers.
High-performance computing is used to develop algorithms for tasks like data collection, signal generation, and trade execution.
Key research areas include machine-readable news, sentiment analysis, and machine learning to enhance performance and address risks, such as those observed during "The Flash Crash" in 2010.
Exploring Machine Learning Methods
Patrik Gabrielsson's research aims to explore various machine learning methods and compare their performance across different scenarios in algorithmic trading.