Thalesians: Mark Salmon - The Importance of Casual Machine Learning in Asset Management
Marriott Hotel Canary Wharf
Wednesday, November 20, 2019 - 07:00 PM
Until: Wednesday, November 20, 2019 - 09:00 PM
(Adjusted for timezone: Europe/London)
Categories
Abstract
Machine learning is currently widely adopted in asset management. Causal methods, which are at the heart of scientific analysis and which have previously been successfully introduced into fields such as genetic research, epidemiology, ecology and climate change, offer powerful ways to enhance predictive analytics by producing forecasts that are more robust and less sensitive to latent confounders. Building causal structures into machine learning processes and the use of counter-factual scenario modelling can help to avoid spurious results and to gain interpretability. Methods such as partial dependence plots, invariant causal prediction and causal forests are presented in the talk.
Brief Bio
Professor Mark Salmon currently teaches Applied Asset Management on the MPhil. Economics and Finance at Cambridge University and is also a Visiting Professor at Imperial College and Director of Research in High Frequency Trading in the Centre for Advanced Financial Engineering (CAFE). He serves as an advisor to Lansdowne Partners Austria and previously served as an advisor to Old Mutual Global Investors (now Merian Global Investors) for 11 years, where he advised the respective Market Neutral Global Equity funds. Between[masked] he acted as Senior Scientist for Brevan Howard David Gorton Systematic Trading, a managed futures CTA. Previously, he advised the Bank of England and the European Commission. His general research interests lie in the theory and application of statistical methods, financial econometrics and the development of trading and predictive analytics including machine learning.
Machine learning is currently widely adopted in asset management. Causal methods, which are at the heart of scientific analysis and which have previously been successfully introduced into fields such as genetic research, epidemiology, ecology and climate change, offer powerful ways to enhance predictive analytics by producing forecasts that are more robust and less sensitive to latent confounders. Building causal structures into machine learning processes and the use of counter-factual scenario modelling can help to avoid spurious results and to gain interpretability. Methods such as partial dependence plots, invariant causal prediction and causal forests are presented in the talk.
Brief Bio
Professor Mark Salmon currently teaches Applied Asset Management on the MPhil. Economics and Finance at Cambridge University and is also a Visiting Professor at Imperial College and Director of Research in High Frequency Trading in the Centre for Advanced Financial Engineering (CAFE). He serves as an advisor to Lansdowne Partners Austria and previously served as an advisor to Old Mutual Global Investors (now Merian Global Investors) for 11 years, where he advised the respective Market Neutral Global Equity funds. Between[masked] he acted as Senior Scientist for Brevan Howard David Gorton Systematic Trading, a managed futures CTA. Previously, he advised the Bank of England and the European Commission. His general research interests lie in the theory and application of statistical methods, financial econometrics and the development of trading and predictive analytics including machine learning.