Optimization in Insurance

Congress Centre Bernardin, Portorož, Slovenia

Congress Centre Bernardin, Portorož, Slovenia

Obala 2, 6320 Portorož
Jasna Prezelj (FAMNIT, FMF, IMFM)

This internal satellite conference is held during the 8th European Congress of Mathematics - 8ECM.

The scope of the conference Optimization in insurance is  to present concrete optimization problems in insurance, from gathering and storing data to risk control. The lecturers come mostly from industry and will present concrete examples they have encountered in their professional work.   The contributions are of particular interest to risk managers and actuaries working in the insurance industry.  The lectures are aimed at ideas how to make companies more competitive.  

Participants are invited to contribute with a poster presentation. 


Organized by:

University of Primorska, with cooperation of  Organizers of the 8ECM 


Slovenian Insurance Association 



    • 09:00 09:50
      How to model and implement a centralized data warehouse to spend more time on analyzing data than gathering it? 50m

      Most of the teams that focus on data-intensive work (analytics, modelling, optimization, etc.) still spend most of the work on data preparation and data cleaning. On the other hand, there is a huge spike in a number of new cloud applications that organizations use to manage their business. Because of all this, the need for data integration and holistic view of all data is even more crucial for managing the business. The presentation will focus on different possibilities of how to model the data warehouse (dimensional, anchor, 3NF form) and how the whole implementation process looks like. With more than 15 years of experience building complex data warehouses in financial institutions, we will share some of the best practices, things we learned the hard way and what are the future trends.

      Speaker: Grega Jerkič (In516ht)
    • 09:50 10:40
      Best practices in data visualization – how to tell data stories using dashboards? 50m

      You have consolidated, cleaned and organised your data. Calculated all the needed metrics and the final step is the presentation of the data. Usually the previous steps are much more demanding, but poor presentation of information will make all previous hard work obsolete. What is most important, how to position it, what kind of data I have, what would be the best chart for my type of data, what is data storytelling, etc.? We will try to answer all this questions with different examples and present the methodology how to create “customer” journey based on your target groups, so that you can learn some of the best practices and create even better dashboards in the future.

      Speaker: Žane Logar (In516ht)
    • 10:40 11:10
      Coffee break 30m
    • 11:10 12:00
      Implementing data science in insurance 50m

      Real-life examples of insurance companies and their path of implementing data science in their business. Where they began and which business problems they covered and what are future challenges and opportunities that will give them an advantage in the market. Why the path was hard and slow and what has changed in recent years will be covered.

      Speaker: Iztok Šerbec (In516ht )
    • 12:00 12:50
      The role of knowledge graphs in insurance 50m

      The lecture we present the use of knowledge graphs in insurance and the algorithms on how to build them, algorithms to inferring missing facts. We focus on healthcare-related knowledge graphs occurring in insurance.

      Speaker: Diego Klabjan (Northwestern university)
    • 12:50 14:20
      Lunch break 1h 30m
    • 14:20 15:10
      Optimization of reinsurance 50m

      Reinsurance is a key risk mitigation tool, particularly in general insurance. An insurer will cede some profit to a reinsurer in order to reduce the risk of its insurance result and make insurance results more stable. Different types of reinsurance can be used for different insurance segments. For each individual reinsurance, type insurer needs to decide how much of the risk will be transferred. Reinsurance optimization is a procedure that tries to find optimal reinsurance structure for each individual insurance segment, so that the optimal balance between the cost and risk transfer is achieved and the expected total insurance result volatility is within the company risk limits.

      Speaker: Bor Harej (Prime Re Solutions)
    • 15:10 16:00
      Guaranteed minimum withdraw benefit 50m

      We will study the problem of the guaranteed minimum withdraw benefit (or GMWB for short) rider pricing by the means of American option pricing in the Lévy setting with a new method. Using wavelet discretization we reduce the problem to a matrix linear complementary problem. This approach with wavelets seems to be helpful in allowing practitioners to decide in advance how accurate a result they want and decide on the way whether to refine the mesh or stop the iteration procedure.

      Speaker: Rok Okorn (Ektimo)
    • 16:00 16:20
      Coffee break 20m
    • 16:20 17:10
      Use of Sub-Weekly Updated Satellite Imagery for Assessment of Damage - A Short Course on Earth Observation Methods and Data 50m

      In recent years the Earth Observation (EO) sector is going through a revolution due to the growth of data volume and quality, a significant part of which is available as open data. The European Copernicus mission provides new data covering the globe every 1-5 days, including both multi-spectral imaging and synthetic aperture radar data, among others. This provides unprecedented insight into the Earth, and is also useful for the assessment of damage due to floods and drought, hail and other extreme weather events. For every event one can get data about the immediate aftermath as well as the data prior, making it possible to estimate the state of the ground before and recovery afterwards. During the talk we will guide users through the basics of EO data and demonstrate how the data can easily be accessed. This will be followed by examples and conclude with a brief introduction into how advanced machine-learning models can be used to analyse the data on a large scale.

      Speaker: Marko Repše (Sentinel Hub)