Deep learning of algorithmic trading strategies & limit order book dynamics

The transcendental advancements over the past decade in automated trading systems and theoretical models of deep learning have led to the rise of machines as the primordial arbiter of decision making in financial markets. Consequently, markets organised around limit order books (LOB) have come to be dominated by algorithmic trading strategies based on objective rules derived form the market microstructure. Traders are increasingly capable of wielding deep learning tools to extract complex abstract hierarchical representations from the plethora of high-frequency financial data available. Despite these manifest changes to the trader ecosystem driven by significant investment in automated technology and algorithms, there remain only fractured frameworks through which regulators, practitioners and academics view this new complex and intricate trader environment. The objective of this thesis is to develop deep learning models capable of defining algorithmic traders in this new environment by the type of strategy the firm is conducting, analysing the impact of these strategies on UL equity market quality and modelling the nexus between algorithmic trading strategies and LOB dynamics.