A Story of Trying to Forecast Extremes: from Statistical Modelling to Commercialisation via Machine Learning

Friday 18th November 2-4pm
University of Leeds, Worsley 9.58

The LIDA Statistical Data Science programme are delighted to welcome Leonid Bogachev for a guest seminar on his application to patent an algorithm.

After a 45minute talk from Leonid, there will be a Q&A and the opportunity for networking afterwards.

Abstract
Credible probabilistic forecasting of future extremes is of paramount importance in many areas, such as environment & weather (floods, air pollution, storms, heatwaves); healthcare (epidemics, hospital admissions, pre-emptive diagnostics); civil planning & management (demography, services); manufacturing & sales (logistics, supply chains, inventories); engineering (materials rapture, durability of constructions & machines); finance & insurance (risk management, financial crashes, insolvency); etc. Statistical methods are well developed for stationary data but this assumption doesn’t hold in many practical situations, where extreme values may strongly depend on the varying environment, so their forecasting presents a big challenge. In the talk, I will outline the approaches to this problem developed by my team, based on a smart combination of statistical models and powerful machine learning tools for online processing of nonstationary data streams. I will also describe recent commercial developments, including a new University startup 4-Xtra Technologies Ltd (with a major focus on financial services) and a patent application for the IP protection.

About Leonid
Education: BSc/MSc Mathematics (Distinction) + PhD Probability/Statistics (Moscow State University).
Research interests & expertise: Probability, Random Processes, Statistical Physics, Statistics, Extreme Value Theory. Over 50 peer-reviewed papers.
Associate Editor:  “Statistics & Probability Letters” (Elsevier); MDPI “Mathematics” (section “Probability & Statistics”).
Former awards: Royal Society Incoming Fellowship (PI/host); Leverhulme Research Fellowship; ZiF Research Group (Bielefeld); Research England/NTI grant “Scalable Machine Learning for Data Stream Forecasting of Extreme Values (4-Xtra)”; EPSRC IAA PoC grant “Commercialisation of the 4-Xtra Opportunity for Online Forecasting of Extreme Values”
Positions (current): Reader in Probability, Department of Statistics, School of Mathematics; Turing Fellow, Alan Turing Institute; Co-Founder & Academic Lead, 4-Xtra Technologies Ltd; Founding Co-director, Artificial Intelligence Horizontal Programme (LIDA)
Register today