Fagseminar skadeforsikring - Pricing, modernization, automation and machine learning
Fagkomité Skade i Den Norske Aktuarforening (DNA) inviterer til seminar om Pricing, modernization, automation and machine learning
Tema for seminaret er prising, og ulike aspekter ved prisingen av forsikringsrisiko. Erlend Willand-Evensen og Are Flataker Viken vil snakke om modernisering av prisingsprosessen og deling av data mellom selskaper. Daniel Faller vil gi en innføring i geografisk prising og diskutere Modifiable Area Unit Problem og spatial smoothing. Vi får også besøk fra Sveits, og Dr Franck Cuypers vil gi en innføring i GLM og Neural Networks før han vil vise hvordan modellering og optimalisering av prising kan gjøres i excel (hvor alle kan deltagerne kan gjennomføre de samme beregningene på egne PCer). Vi får også besøk fra Universiteter i Stockholm med Mathias Lindholm, som vil fortelle om utfordringer med å unngå indirekte diskriminering ved bruk av komplekse modeller.
|
08:50-09:00 |
Velkommen Sandra Brännstam, leder i Fagkomité Skade |
|
|
09:00-10:00 |
Pricing and proxy discrimination As data collection and computing power have become cheap, powerful data-driven statistical machine learning methods for prediction have received interest in many application areas, not least in insurance. Common to many of these methods is that they tend to be difficult to interpret; "black box" methods. As a result, questions about implicit, "proxy", discrimination and how this can be avoided have been raised. However, this is a problem regardless of the prediction method used. In this presentation, I will discuss different aspects of this relating to so-called discrimination-free insurance pricing, algorithmic fairness and causality.
|
|
|
10:00-10:10 |
Pause |
|
|
10:10-11:10 |
Geographical Rating Factors In this presentation, we delve into the importance of risk classification in insurance and explore the role of geographical factors in determining the frequency and severity of potential risks. We will establish a common language to discuss geographical insurance risk and examine how location may explain variations not accounted for by individual policyholder attributes. By reviewing relevant papers on premium rating with geographic information and discussing the Modifiable Areal Unit Problem (MAUP), we will emphasize the challenges of drawing inferences from aggregated data. Finally, we will share insights from FCG's research on portfolio-tailored rating regions, highlighting the significance of spatial smoothing and addressing the challenges posed by data-sparse portfolios. Presentasjonen vil bli holdt på engelsk. Daniel Faller, Analyst in the Actuarial and Advanced Analytics Team at FCG |
|
|
11:10-11:15 |
Pause |
|
|
11:15-12:30 |
Simplifying and understanding machine learning In this hands-on seminar, after an introduction to the basics of Linear Models, Generalized Linear Models and Neural Networks, we will implement their algorithms in Excel and apply them to the pricing of a simple accident insurance product. Based on the obtained rates, we will then explore how different insurers can optimize their profitability and market share depending on the quality of their pricing engine. I will distribute the Excel template beforehand, so feel free to bring your laptop and do the calculations with me while I demonstrate. Presentasjonen vil bli holdt på engelsk. Dr Frank Cuypers, Prime Re Solutions |
|
|
12:30-13:20 |
Lunsj |
|
|
13:20-14:00 |
Forts. Simplifying and understanding machine learning Dr Frank Cuypers, Prime Re Solutions |
|
|
14:00-14:05 |
Pause |
|
|
14:05 – 15:05 |
Modernization of insurance pricing - a managers perspective The world of insurance pricing has been changing rapidly over the past years. Possibilities – and challenges – seem endless. How can an old incumbent ensure a future-proof pricing operation, and at the same time stay deliver a stable P&L. Erlend Willand-Evensen, Gjensidige |
|
|
15:05 – 15:15 |
Pause |
|
|
15:15-16:15 |
Pricing department collaboration and data sharing between companies We explore how pricing departments in insurance companies, both big and small, can learn from each other. We also ask the question of how Analytika has pulled off having a team of pricing actuaries working as consultants, considering pricing actuaries general allergy to consultancy babble and circumlocution. Lastly, we talk about the data sharing company Aeda, and how some Norwegian insurance companies boost their analytical insight by sharing data with each other. Are Flataker Viken, Analytica |
|
|
16:10-16:15 |
Oppsummering |
|
Seminaret gir 7 poeng for internasjonalt godkjent aktuar.
Påmelding
Påmelding er bindende og gjøres nederst på denne siden senest 4. mai 2023.
Ved spørsmål kontakt admin@aktfor.no
Det tas forbehold om endringer i programmet.
Mathias Lindholm
Mathias Lindholm är docent i matematisk statistik och arbetar som lektor vid Stockholms universitet och forskar primärt inom försäkringsmatematik. Innan Mathias gick tillbaka till akademin på heltid har han arbetat flera år som aktuarie och kvantitativ analytiker med frågor som rör riskkontroll, strategisk allokering och ALM, inklusive metodutveckling.
Dr Frank Cuypers
Frank Cuypers is a nuclear engineer, with a M.Sc. in nuclear physics and a Ph.D. in theoretical physics. He comes with a prominent scientific track record in modelling complex systems and an extensive lecturing experience.
As a former Swiss Re Executive and Chief Actuary of Zurich Re, he has gained broad expertise in most actuarial disciplines, which he now deploys to advise leading insurers & reinsurers with bespoke technical solutions. Regularly numerous companies and regulators appoint him to develop and validate risk and solvency models, and he has pioneered individual claims development methods with machine learning techniques.
Frank is a fully qualified actuary of the German and the Swiss Actuarial Associations. He served until 2022 on ASTIN’s Board, which he chaired for 5 years, and he founded the CPD Committee of the Swiss Actuarial Association, which he presided for 10 years. He is also a shareholder and actuarial engineer of Prime Re Solutions, an actuarial consultancy.
Daniel Faller
Daniel Faller is an analyst in the Actuarial and Advanced Analytics Team at FCG, holding an M.Sc.Eng in Risk Management with a specialization in Mathematical Statistics. He has led projects in risk premium forecasting, pricing, geographic risk strategy, operational risk management, and ESG strategy. Daniel's understanding of risk management and statistical analysis helps FCG provide actionable insights to its clients.
Erlend Willand-Evensen
Erlend Willand-Evensen trained as an economist at NTNU and UIO and has close to 20 years of experience from Product, Price and Underwriting roles with some of the largest insurers in the Nordic markets. For the past seven years he has held the position of SVP Head of Pricing at Gjensidige.
Are Flataker Viken
Are Flataker Viken holds degrees in physics and statistics from the University of Bergen. He was a former senior pricing actuary in Gjensidige, before starting in Analytika in 2019. He is also the CEO and co-founder of Aeda, a company focusing on sharing data between companies.
Are F. Viken snakket om hvordan forsikringsselskaper kan lære av hverandre og hva de kan fokusere mer på for å få mer effektive og bedre prisingsavdelinger. De beste selskapene jobber prosjektbasert for å komme til høyere og mer stabile prestasjonsnivå, og Are snakket om hvordan Analytika samarbeider med selskapene rundt dette. Til sist ble aktuarforeningen fortalt om hvordan Aeda hjelper forsikringsselskaper å dele data med hverandre på en rettferdig måte
Sted
Bryggetorget 3, 0250 Oslo, Norway