Non-Life Pricing & Profitability Analysis Using ML Techniques with R Applications

Event Type
Web Session
European Actuarial Academy GmbH in cooperation with the Eesti Aktuaaride Liit.

Tuesday 1st December and Wednesday 2nd December

Start time: 7.00 am GMT

End time: 7.00 am GMT


Announcement from the European Actuarial Academy:  Non-Life insurance is facing many challenges ranging from fierce competition on the market or evolution in the distribution channel used by the consumers to evolution of the regulatory environment.
Pricing is the central link between solvency, profitability and market shares (volume). Improving pricing practice encompasses several dimensions:

  • Technical: is our pricing adequate to cover the underlying cost of risk of my policyholders and the other costs we are facing? Which are the key variables driving the risk? Are they adequately taken into account in our pricing? What’s the impact of the claims history of my policyholder on its expected risk? In which segment are we profitable and in which are we not profitable?
  • Competition: at what price will we attract the segments that we target and price out those that we do not want? Is the positioning of our competitors influencing our pricing practice and our profitability? What’s my position with respect to my competitors in term of pricing? What are the segments in which I am well positioned and the segments where I am not well positioned?
  • Elasticity: what price (evolution) are our existing customers prepared to accept? Does the sensitivity to price evolution depend on the profile of my customer?
  • Segmentation: is our segmentation granular enough for our purposes?

The aim of this web session is to present some advanced actuarial/statistical techniques used in non-life pricing, competition analysis and profitability analysis. The web session focuses on some practical problems faced by pricing actuaries and product managers and presents some new techniques used in non-life pricing in order to open new perspectives for product development (competition analysis, profitability analysis,…)

The web session is developed for non-life actuaries or statisticians but also for managers working in product development or risk management departments. Participants should ideally have basic knowledge of non-life pricing.

Attendees are encouraged to have a laptop computer with R installed as well as some useful packages (all the information will be provided after subscription). A basic knowledge of the R software is useful.

Technical requirements: Please check with your IT department if your firewall and computer settings support web session participation (the programme GotoTraining/GotoWebinar is used for the web session).


The web session will alternate between methodological concepts, practical examples and case studies in order to ensure a comprehensive understanding of the techniques presented.

The participants will be requested to look at 4 e-learning modules (of around 30 minutes each) presenting the basics of machine learning before the web session. The access to these e-learning modules will be granted up to the end of the seminar.
These 4 e-learning modules are

  • Introduction to Machine learning
  • Supervised learning (parts 1 and 2)
  • Unsupervised learning

The case studies will be performed by the participants with the R software. An individual support can be available during the case study sessions through individual Teams meeting with the  speakers.

Click here to make a reservation. The registration fee is € 640 plus 16% VAT until 1 October 2020. After this date, the fee will be € 790 plus 16% VAT


Tuesday, 01 December 2020

  • Data Selection, Pre-Analysis and Feature Selection (data quality, pre-treatment, missing values, feature engineering and feature selection)
  • Overfitting and cross-validation
  • Case study: Data Analysis and filtering
  • 8.45 am - 9.00 am: Coffee Break
  • Reminders and Q&A about supervised learning models
  • 10.30 am - 11.30 am: Lunch
  • Example: Fitting a regression tree and random forest on frequency
  • Presentation: Case Study: Regression tree and random forest model adjustment for frequency and cost
  • Presentatition Case study: Reminders and Q&A about supervised learning models
  • Participants work by themselves on the case studies (individual support available thanks to Teams meeting)
  • Correction of the case studies and closing of the day

Wednesday, 02 December 2020

  • Interpretability of Machine learning techniques
  • Case study: Features selection, partial dependence plot and Shapley Value
  • Case study: Application of GBM method to highlight interactions
  • Participants work by themselves on the case studies (individual support available thanks to Teams meeting)
  • Correction of the case studies
  • 10.00 am- 11.00 am: Lunch
  • Reminder and example about unsupervised machine learning
  • 12.00 pm - 12.15 pm: Break
  • Profitability and Competition analysis (profitability and positioning assessment, reverse engineering of competitors prices)
  • Example: profitability analysis with regression trees
Samuel Mahy (Reacfin) and Xavier Maréchal (Reacfin)
Biographical details

Samuel Mahy (Reacfin)
Samuel graduated as a Master in Engineering (Applied Mathematics) with an additional minor in Economy and holds a Master in Actuarial Sciences, as well. He is a qualified actuary of the Institute of Actuaries in Belgium (IA|BE) and involved in the Reinsurance and Non-Life Workgroup of the IA|BE. He is the Head of the Non-Life Center of Excellence at Reacfin. Samuel has been active 5 years in the reinsurance sector where he was involved in reinsurance pricing model developments. At the same time he was also the main responsible of the UK market portfolio profitability follow-up. Samuel joined Reacfin in June 2010 as a specialist in Non-Life Insurance and Reinsurance and he has acquired a sound knowledge of Solvency 2 frameworks (Non-Life, Health). As a director, he is involved in various missions as in the modelling, implementation and validation of pillar I deliverables (standard approach and (Partial) Internal models), reinsurance optimization, model documentation, non-life pricing model development for several lines of business, etc..

Xavier Maréchal (Reacfin)
Xavier is founder and CEO of Reacfin. Xavier is one of the co-authors of “Actuarial Modeling of Claim Counts: Risk Classification, Credibility and Bonus-Malus Systems” (Wiley, 2007). Xavier has obtained different academic degrees as Master in Engineering (Applied Mathematics), MSc. Actuarial Sciences and MSc. Management. Xavier is a qualified actuary of the Institute of Actuaries in Belgium (IA|BE). Xavier has extensive experience in the actuarial field obtained during his 15 years as a principal consultant for many national and multinational insurance companies. He has gained a complementary experience in various fields going from Non-Life ratemaking and provisioning to health modeling and ALM. After several years of intensive modeling activities in health, non-life and ALM, Xavier works now as reviewer and mentor for consultants. He performed several validation assignments and holds the actuarial function for a health insurance company.