Deep learning for Insurance and Finance

Teacher(s)

Hainaut Donatien;

Language

French
> English-friendly

Prerequisites

Knowledge of basic concepts in statistics and probability calculation as well as programming, at the course level of the FSA1BA, INGE1BA, MATH1BA programs or the access minor in statistics, actuarial sciences and data science.

Main themes

Artificial neural networks, deep learning, auto-encoder, LSTM, convolution networks, pricing and forecasting

Learning outcomes

At the end of this learning unit, the student is able to :
1	Program a regression/classification artificial neural network for pricing or forecasting in insurance/finance Choose the calibration algorithm, the optimization criterion, as well as the constraints most appropriate to the data Implement a neural network reducing the size of a dataset Using a network to model time series

Content

• Neural networks (NN), general
• NN for insurance and credit risk: deviance and personalized loss function, bias regularization.
• NN and high dimension: ridge and lasso penalization as well as “embedding layers”
• Bias-variance: bootstrapping, randomization drop-out
• Interpretation of models: PDP, ICE, feature importance, LIME, SHAP
• Neural autoencoders and variational autoencoders
• Time series forecasting with recurrent networks and LSTM
• Regression with convolution network

Teaching methods

• Reading with slides
• Programs in Python (KERAS & TENSORFLOW)
• Case studies

Evaluation methods

Students will prepare an individual report in which the methods seen during the assigned readings are applied to a real-world dataset. Note that the professor reserves the right to question students orally on the content of their work. Where applicable, the use of AI will comply with the guidelines for the responsible use of generative artificial intelligence within the Faculty of Science.

Online resources

Moodle website

Bibliography

Denuit M., Trufin J. , Hainaut D. 2019. Effective statistical learning III : neural networks and extensions. Springer actuarial lectures notes.

Faculty or entity

> LSBA