Description: This article gives a brief overview of ways to extend clustering analysis beyond basic methods such a k-means

Description: Actuaries can empower their understanding of policyholder behavior with third party data. Modeling policyholder behavior should start from developing knowledge of who they are, instead of actuaries assuming what a rational behavior should be.

Description: Given the limits of classical actuarial modeling and forecasting techniques, this newsletter contains articles discussing several ways to supplement your toolset.

Description: The first annual SOA Predictive Analytics Symposium was held in Chicago and exceeded all expectations in attendance, presentation quality, and relevant content.

Description: The author discusses the two-part presentation on complexity science tools at the SOA Annual Meeting October 2010. He explains that presentations went back to basics with explanations about the difference between modeling tools used to employ deterministic methods for classical actuarial models.

Description: The author discusses expectations for incoming actuarial science students, emphasizing that a solid grasp of algebra and traditional mathematical skills is not enough. In addition, the author states that today’s business environment dictates that students be well versed and that their backgrounds include familiarity with complexity science.

Description: Can the aggregate results from a group be used to forecast something as subjective and judgmental as one single person’s opinion? Forecasting of this type commonly uses a set of techniques known as collaborative filtering which is discussed at length in this article.

Description: Machine learning models are not black-boxes; these models can easily be interpreted using tools like SHAP, LIME and Anchor. These tools can facilitate the understanding of the model and the business problem. We can used these tools to build better models and communicate how predictions are generated. This article demonstrates their use in a mortality model.

Description: This article explores the use of Generalized Additive Models and Multivariate Adaptive Regression Splines to model nonlinear relationships in a linear response model. These models use supervised methods to determine the optimal shape of the predictors in the model. Users are provided an example use case so that they may get started right away building models to explore complex relationships in new ways.

Description: Spark offers a way to get some dramatic speed and scalability advantages over the Map-reduce methodology whenever you are doing iterative data processing and particularly when you want interactive data analysis capability. Find out why “lazy execution” can be a desirable characteristic when dealing with big data.