Category Archives: Research

Personalization as a promise: Can Big Data change the practice of insurance?

Our recent article, “Personalization as a Promise: Can Big Data Change the Practice of Insurance?” with Larence Barry just got published in “Big Data & Society” (but a working paper version is still available online)

The purpose of this paper is to measure the impact of technologies from the Big Bang. data on thehe pricing ofhe products car insurance. The first part describes how the aggregated view buildsuit by statistics enables highlighting invisible regularities at the individual level. Despite a very granular segmentations in automobile insurance, the approach remained classificatory, hypothesizing the risk identity of individuals from the same class. The second part highlights the reversal of big data-induced perspective in the’analysis ofgiven ; awith theur volume and the new algorithms, the aggregate viewpoint is questioned.. The hypothesis of class homogeneity is becoming increasingly difficult to test. maintain, especially since predictive analysis boasts the ability to predict the rs results at the individual level. The third part is studying the’influence of telematics boxes able to import the new pinsurance aradigm automobile. However, a reading of the most recent research articles on a pricing automobile including this new monter that the epistemological leap, at least for now, has not taken place.

COVID19 pandemic control: balancing detection policy and lockdown intervention under ICU sustainability

After almost two months, with Romuald Elie, Chi (Tran Viet Chi) and Mathieu Laurière, we finally have a draft of the paper related to our recent work, entitled COVID-19 pandemic control: balancing detection policy and lockdown intervention under ICU sustainability” (available on HAL ArXiv and MedrXiv)

Our model is an extention of the classical SIR model, but more realistic for the COVID-19 (pronounded “cider”)

We use scenarios to see the impact on various quantities

but also optimal control to see the best strategy, when it comes to lockdown, and testing. All comments are welcome…

 

Quelles limites pour les algorithmes d’apprentissage ?

Avec Michel Denuit, on vient de publier un court article, quelles limites pour les algorithmes d’apprentissage ? Une version en anglais devrait paraître dans un ouvrage collectif dans les semaines à venir.

Une étape essentielle du processus de tarification est le choix de l’algorithme qui permettra de calculer la prime à partir de données passées. En assurance, les modèles prédictifs sont partout : il peut s’agir de calculer la prime demandée pour couvrir les dommages causés à un bien, ou encore d’un score interne de suspicion de fraude sur un sinistre. Ce papier discute des limites de l’application des algorithmes prédictifs en assurance en revenant sur les notions de segmentation des risques et d’hétérogénéité.

Telematics: a revolution ?

Tomorrow morning, Laurence Barry will present some joint work at the Online International Conference in Actuarial science, data science and finance, organised by colleagues in Lyon. The paper, “Personalization as a Promise: Can Big Data Change the Practice of Insurance?” is online, as a Working Paper of the PARI chair,

The purpose of this paper is to measure the impact of technologies from the Big Bang. data on thehe pricing ofhe products car insurance. The first part describes how the aggregated view buildsuit by statistics enables highlighting invisible regularities at the individual level. Despite a very granular segmentations in automobile insurance, the approach remained classificatory, hypothesizing the risk identity of individuals from the same class. The second part highlights the reversal of big data-induced perspective in the’analysis ofgiven ; awith theur volume and the new algorithms, the aggregate viewpoint is questioned.. The hypothesis of class homogeneity is becoming increasingly difficult to test. maintain, especially since predictive analysis boasts the ability to predict the rs results at the individual level. The third part is studying the’influence of telematics boxes able to import the new pinsurance aradigm automobile. However, a reading of the most recent research articles on a pricing automobile including this new monter that the epistemological leap, at least for now, has not taken place.

De la démarche scientifique en période de crise

Dans une conférence donnée le 13 février 2020[i], intitulée contre la méthode, Didier Raoult affirmait « moi je n’ai jamais fait d’essais randomisés […] faire ça sur des maladies infectieuses, ça n’a pas de sens ». Cette vision était reprise dans une tribune plus détaillée, où face à « la méthode » (et « aux mathématiques »), Didier Raoult défendait (ce qu’il appelait) « la morale [et] l’humanisme » du serment d’Hippocrate. Comme il le rappelle, faire des groupes de contrôle, c’est « dire au malade qu’on va lui donner au hasard soit le médicament dont on sait qu’il marche, soit le médicament dont on ne sait pas s’il marche » (Raoult (2020a, 2020b)). Alors que cette méthode d’expériences randomisées est aujourd’hui saluée dans toutes les disciplines – comme le rappelle le prix Nobel d’économie attribué en 2019 à Esther Duflo, Michael Kremer et Abhijit Banerjee – comment un chercheur peut-il prendre une telle position, aujourd’hui ? Continue reading De la démarche scientifique en période de crise

Reinforcement Learning in Economics and Finance

With Romuald Elie and Carl Remlinger we recently uploaded on ArXiv a paper on Reinforcement Learning in Economics and Finance

Reinforcement learning algorithms describe how an agent can learn an optimal action policy in a sequential decision process, through repeated experience. In a given environment, the agent policy provides him some running and terminal rewards. As in online learning, the agent learns sequentially. As in multi-armed bandit problems, when an agent picks an action, he can not infer ex-post the rewards induced by other action choices. In reinforcement learning, his actions have consequences: they influence not only rewards, but also future states of the world. The goal of reinforcement learning is to find an optimal policy — a mapping from the states of the world to the set of actions, in order to maximize cumulative reward, which is a long term strategy. Exploring might be sub-optimal on a short-term horizon but could lead to optimal long-term ones. Many problems of optimal control, popular in economics for more than forty years, can be expressed in the reinforcement learning framework, and recent advances in computational science, provided in particular by deep learning algorithms, can be used by economists in order to solve complex behavioral problems. In this article, we propose a state-of-the-art of reinforcement learning techniques, and present applications in economics, game theory, operation research and finance.