Join us in our upcoming webinar on: 
The use of autoencoders for training neural networks with mixed categorical and numerical features

21 SEP 2023 | 10 AM EST
10:00 AM (EST) 16:00PM (CEST) 

We focus on modelling categorical features and improving predictive power of neural networks with mixed categorical and numerical features in supervised learning tasks. The goal of the paper is to challenge the current dominant approach in actuarial data science with a new architecture of a neural network and a new training algorithm. The key proposal is to use a joint embedding for all categorical features, instead of separate entity embeddings, to determine the numerical representation of the categorical features which is fed, together with all other numerical features, into hidden layers of a neural network with a target response.
In addition, we postulate that we should initialize the numerical representation of the categorical features and other parameters of the hidden layers of the neural network with parameters trained with (denoising) autoencoders in unsupervised learning tasks, instead of using random initialization of parameters. Since autoencoders for categorical data play an important role in this research, they are investigated in more depth in the paper. We illustrate our ideas with experiments on a real data set with claim numbers, and we demonstrate that we can achieve a higher predictive power of the network.

Łukasz Delong | Speaker 
Łukasz Delong is working at SGH Warsaw School of Economics. He has a PhD in mathematics, a habilitation degree in economics and the professor title in economics and finance. Łukasz is an actuary with license no. 130 issued by the Polish Financial Supervision Authority, the head of the Examination Committee for Actuaries at the Polish Financial Supervision Authority, a board member of the Polish Society of Actuaries and a board member of AFIR-ERM Section of the IAA. He is an editor of ASTIN Bulletin - The Journal of the IAA and an associate editor of European Actuarial Journal. Łukasz is an author of numerous scientific papers on insurance mathematics, financial mathematics and probability. His scientific research includes different areas of actuarial mathematics with emphasis on stochastic modelling of financial risks and machine learning in insurance.

Brian Fannin | Session Moderator
Brian Fannin is a research actuary at CAS. He is also the founder of PirateGrunt LLC, a boutique consulting firm based in Durham, North Carolina specializing in predictive modeling in the property casualty markets. Fannin has been an Associate of the CAS since 2002 and a Certified Specialist in Predictive Analytics (CSPA) through The CAS Institute since 2017