Date/Time
Date(s) - 01/09/2023
10:00 am - 11:30 am
Categories
Mastering Insurance Pricing
Presented by Quantee, www.quantee.ai
The course is delivered in four live online sessions, each of 1.5 hours, 10am – 11.30am UK time, on Friday.
Session 1 – Mastering pricing processes – 1st September 2023
Session 2 – Advanced data analysis – 8th September 2023
Session 3 – Building and maintaining generalized linear/additive models – 15th September 2023
Session 4 – Utilizing (explainable) AI in pricing – 22nd September 2023
The course is aimed at a technically astute audience interested in general insurance pricing and focuses on presenting state-of-the-art techniques. The Quantee platform is used. Further details of the sessions is as follows.
Session 1 – Mastering pricing processes
Building a live pricing process on the platform. An overview of a pricing process is as follows
a. Describe each constituent in a sample pipeline/presentation
b. Explain pricing sophistication journey, (i) a simple rating table case, (ii) linear models, (iii) non-linear models, (iv) Briefly description of price optimization
c. Describe the process of adding loadings
d. Design the final product.
Session 2 – Advanced data analysis
Start with introduction to methods that will be covered and what is meant by data analysis in terms of actuarial processes, then do explanatory data analysis.
a. Import a raw dataset
b. Explain the available charts
c. Introduce geostructure and plot relevant data
d. Introduce lookup dataset, eg to include inflation rates
e. A quick overview of an open-source integration in a form of plugin
f. Finish with a clean dataset ready for modelling purposes
g. Mention how dashboards and Data Visualizations techniques could be used for Portfolio Monitoring purposes
Session 3 – Building and maintaining generalized linear/additive models
Start with theoretical background on GLMs/GAMs – cover statistical and mathematical details, history and use cases.
a. AutoGLM mode will be used for a benchmark
b. A relatively small dataset shall be used so that each model version can be re-calculated quickly. Starting with a simple model, layers of complexity will be added iteratively
c. The goal is to build a model from scratch that would improve or at least match theAutoGLM model.
Session 4 – Utilizing (explainable) AI in pricing
a. Provide background on ML techniques, eg academic theory, use cases, why NNs not great right now and the need for balance of accuracy & transparency
b. Build Gradient Boosting Method model
c. Compare to GLM using basic techniques, (i) target vs. predicted, and (ii) double lift chart
d. background of XAI
e. Present XAI dashboard with GBM vs. GBM comparison
Bookings
Bookings are closed for this event.