Time-oriented statistical machine learning, with applications to electronic health-care, social networks and finance

schedule le mardi 05 décembre 2017 de 16h30 à 18h00

Organisé par : Bastien Fernandez, Nicolas Fournier, Sandrine Péché

Intervenant : Stéphane Gaïffas (LPMA)
Lieu : Salle 0011, Sophie Germain (Université Paris Diderot)

Sujet : Time-oriented statistical machine learning, with applications to electronic health-care, social networks and finance

Résumé :

This talk gives an overview of contributions in the field of statistical machine learning for time-oriented data science problems. The main applications are in the field of health-care, social networks and finance. We will introduce some new statistical techniques, based on a mathematical tool called « point process », in particular Hawkes processes, and « self-control case series » models in the context of health-care. Will we also give a short description of an ongoing project with the French social security, where we apply such tools on their electronic health-care database (one of the largest in the world).

We will give an overview of the key mathematical concepts involved in this, some of the theoretical statistical guarantees derived for our algorithms and we will illustrate them on large-scale datasets.