As actuaries adopt increasingly complex data science techniques, we must always remember that every model corresponds to specific business problems and solutions. Therefore, the best model depends not only on statistical support, but also on business needs and goals. Often, the model with the best test statistics is not the best model for the business and the ability to translate data science results to business outcomes is an essential skill for actuaries.
In this presentation, we will share the results of modeling exercises and ask the audience to translate these charts into clear potential business outcomes. We will also share examples of where the output of a model should be overridden due to actuarial considerations and ask the audience for suggestions. Additionally, we will share good and bad questions to ask about a model to answer specific actuarial questions.
To encourage audience participation, this session will not be recorded.
Learning Objectives:
Draw clear connections between traditional model output and business results.
Identify situations where actuarial considerations should override a statistically optimal selection.
Evaluate how ideal or expected modeling results may differ between modeling perils or coverages.
