2022 Webinar - Elastic Nets, Modeling Applications in Insurance Pricing - April 21
Elastic nets are an innovation on traditional approaches to actuarial pricing models. GLMs have been a standard tool for building insurance pricing models for the last two decades. They have some known limitations, such as granting full credibility to the parameter estimates and being prone to overfitting on segments where data is sparse. Elastic nets introduce a penalty term, the effect of which is to shrink the model parameters toward or to zero. This has the combined effect of limiting the model’s response to outliers and eliminating variables completely from the model when they lack statistical stability. The result is often a more accurate pricing model.
The webinar will include a discussion of foundational elements of elastic nets. An open source demo will showcase the ability of elastic nets to remove noise variables and the selection of appropriate hyperparameters using cross validation. Viewers will see some of the standard model output generated by elastic net fitting software.
- Understand the difference between a traditional GLM and elastic net
- Understand the difference between the types of penalty terms and their behaviors
- Know how to select “hyperparameters” for elastic net models
- Know what business problems can benefit from elastic net regression
- Become familiar with standard model output and know how to interpret it
CAS provides closed captioning generated by software applications for the convenience of viewers. CAS does not review for accuracy of the information that appears in a closed caption, and such information (including any foreign language translation and transcription displayed in a closed caption) may not be relied upon as an accurate or complete record of CAS programming. CAS makes no representations or warranties, and expressly disclaim any responsibility or liability with respect to, any errors or omissions in, or the accuracy, reliability, timeliness or completeness of, any information that appears in a closed caption.