ASTIN Working Party Looking for Volunteers
New ASTIN Working Party is seeking volunteers. The framework of this study is Non-Life reserving. The aim of this working party is to demonstrate the effect of combining Machine Learning techniques and traditional ones, in particular time series and survival analysis methods, on the accuracy of Machine Learning forecast power.
To reach this purpose, the working party plans to:
- Gather Data of a specific Non-Life branch and build a clean data set ;
- Apply classical reserving methods in order to understand the global structure of the portfolio ;
- Identify :
- time series, survival analysis techniques to use, and at which step they have to intervene
- Machine Learning methods to use (type, parameters, limitations, …)
- Build a R program to apply the different methods identified and comment each step ;
- Analyze the impact of using traditional methods in parallel with Machine Learning in reduction of the forecast error.
The works will mainly be realized in R.
Deadline to reply: October 31.
CAS members who are interested should send an email with a short resume to the following address: MLTMS@actuaries.org.