Data-driven methods for timeseries modelling across asset classes

点击翻译按钮获取中文摘要

My current research sits at the intersection between engineering and mathematical sciences, aiming to improve the modelling of cross-asset market microstructure in equity and exchange traded fund (ETF) markets. This multidisciplinary approach leverages techniques from modern statistics and machine learning to obtain insights from multi-terabyte scale, low signal to noise timeseries datasets. This research will provide the first systematic analysis of trade co-occurrence between equities and ETFs. Trade co-occurrence captures the information content of trades occurring in a short time proximity of each other. By performing the first systematic analysis of trade co-occurrence between equities and ETFs, we aim to address fundamental questions in market microstructure such as identifying arbitrage flow and detecting lead-lag relationships. The practical applications of this research are of interest to both regulators as well as many other market participants such as high frequency trading firms (HFTs), market markets (MM) and hedge funds. For instance, the understanding of price formation mechanisms is vital to regulators who seek to understand the flow of information and systemic risk within a market. Analysis of lead-lag relationships is extremely important to market participants such as HFTs and MMs who seek to optimally provide liquidity across a broad set of asset classes.To maximise the impact of my research, I am collaborating with senior quantitative researchers from Man Group. Man Group is the world's largest publicly traded hedge fund, and this partnership will ensure that our research objectives and methodologies align with the practical needs of industry. The technical expertise and guidance from industry practitioners will allow us to maximise our research impact.This research offers multiple avenues for further exploration. The methodology is general enough to be adapted to other asset classes such as options and futures. We aim to extend our framework to go beyond pairwise interactions in asset classes and move to a more general graph-based framework.This research aligns with the EPSRC's objectives of fostering innovation in mathematical sciences and engineering. By leveraging innovative statistical methods and machine learning, we aim to advance understanding in market microstructure across asset classes. The collaboration with industry stakeholders like Man Group underscores its relevance to real-world challenges, fulfilling the EPSRC's mandate for impactful and practical research.