­

­

­

­

EAA Web Session: "Machine Learning and Anomaly Detection" on 1 April 2022
Interview with our speaker Michael Leitschkis

­

­

­

­

­

­

­

­

­

­

­

­

­

­

­

­

Michael, why is outlier detection a relevant topic for actuaries today?
With actuarial models becoming more and more powerful, actuaries have to process much larger volumes of input data than ever before. Thousands and thousands of inputs cannot be manually inspected - and if bad data do enter calculations undetected, then we are facing the "garbage in, garbage out" perspectives. 
Therefore, actuaries need state-of-the art outlier detection capabilities in order to make optimal use of their models. 

Machine Learning surely sounds great, but is it really needed for outlier detection purposes?
Of course, simpler methods also exist - such as Tukey criterion - such approaches will be introduced in this web session before moving on to more complex methods. The danger with using a too simple approach is that the user may end up discarding some useful data while not detecting relevant outliers. 

Is prior expertise required to follow this web session?
Short answer: No. 
Long answer: We introduce all necessary theoretical concepts before applying them to practical applications. Attendees will be familiar with different models to address different business questions in different markets, and yet they will be facing the common challenge of how to automatically detect outliers in their data.

­

­

­

­

­

­

­

­

­

­

­

actuarial-academy.com

­