Algorithmic decision making in financial markets

Thesis

Machine learning's prowess for automatic pattern recognition at scale is meaningfully reshaping every branch of science. From astronomy to vision, web analytics to medical diagnostics, every data-intensive field is harnessing the potential of modern AI techniques. Though not commonly viewed through the same lens, finance is very much at the forefront of the data revolution. Financial markets present one of the most complex, noisy environments for machine learners: a vast range of factors - not all readily quantifiable - may impact a financial time series, and the relative salience of market variables may evolve through time. The aim of this thesis is to investigate algorithmic frameworks for the challenging decisions faced by liquidity takers (the `buy side') and market makers (the `sell side'), the primary agents in financial markets. By extension, we also consider the behaviour of influential external agents such as regulators, whose actions affect the information landscape for buyers and sellers alike. This thesis deploys recent advances in machine learning to provide rational, data-driven tools to promote market efficiency in the areas of price discovery, liquidity provision and financial regulation.

Authors: Ghoshal, S

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English