Recent Posts

More Posts

This blog is a static hugo blog, generated with the amazing R package blogdown. At first shot, i decided to locate the working documents in a different directory as the published hugo site, allowing me to separate commits to the sources from commits of compiled files. But thoose two repo were a problem : things were not centralised enough for me, and i wanted to push on github only the files i’m working on : commiting compiled files is not realy a good practice.

CONTINUE READING

After working on a bootstrapping framework for the Mack model, with a one-year point of view and with several triangles to bootstrap jointly, i decided to put some of my code into a litle package, mbmcl. You can install it with : We’ll also load some conveniance packages, namely magrittr/dplyr/purrr/tibble/tidyr. library(magrittr) library(dplyr) library(purrr) library(tibble) library(tidyr) Then load some triangle data, for exemple the ABC triangle from the ChainLadder package. For the purpose of this exposition, we need several triangles of same size, lets create dummy triangles, and look at mack’s results on them :

CONTINUE READING

My actuarial thesis got published online there This work took me a little more than one year to do, an was dealing with non-life reserving in solvency 2 context for the french decenial insurance contracts. Here’s the abstract : After having specified the specificities of the French builder’s insurance and analyzed the problems posed by additional specific reserves to this line of buisiness, we recall basic models, deterministic and stochastic, used in non-life insurance.

CONTINUE READING

Introduction Suppose you have a dataset, and you are narowing possible machine learning models to 2 or 3 models, but you still cant choose which you want : Will the benefit of understandability from my CART cost me too much compare to a random forest or some bootsting ? Well you dont necessarily have to choose : juste agregate the models you have to make a better one. Typicaly, if you have models that dont uses the same features of the dataset, or give very different ansewrs but are still all good in term of a pre-selected metric (let’s say RMSE for regression, area under ROC for classification), ensembling them could be a good idea.

CONTINUE READING

Introduction The log-normal model Generating dummy dataset. Checking model hypohtesis. Log-normality of \(y_t\) Linearity between \(y_t\) and \(x_t\) Results from the model References Introduction Under Solvency 2 framework, insurance compagnies can calculate undertaking specific parameters to modify their application of the standard formula, as dictates Commission-Européenne (2014) . One of thoose calculations methods uses a log-normal model that’s quite interessante to analyse. We will here analyse this model in a mathematical sense, and then use it on simulated data (i’m not an insurence compagny !

CONTINUE READING

Projects

empCop R package

A package containing empirical checkerboard copula and derivative copula’s S4 classes to work with thoose models in R

Actuarial Master Thesis : Reserving and reserving risk in french builder’s insurance

After having specified the specificities of the French builder’s insurance and analyzed the problems posed by additional specific reserves to this line of buisiness, we recall basic models, deterministic and stochastic, used in non-life insurance. Having transcribed the main models in a larger mathematic framework, we present new models shaped for the tri-dimentional issue of French construction insurance, including the estimation of variability in every point of view. Then, through several different approaches, we derive Solvency II reserve risk estimators in those models, and conclude with an analysis of these estimators on a certain portfolio, shaped toward the standard formula.

mbmcl - Multi-boot Mack chain-ladder

A simple package that implement a one-year bootstrap in the Braun model, with extension to a mutli-year reserving case (french decenial insurance). Work done in the framework of my master thesis.

EstimateCopula Shiny App

A shiny app that estimate and give some statistic about a bivariate copula given a bivaraite dataset.

Estimation of PDD

Some code do estimate a PDD in certain solvency II settings, corresponding to a Numerical technics exemple in insurence.