Description: The advent of machine learning algorithms has taken predictive analytics and risk scoring to a new level. From gradient boosting machines to k-nearest neighbors regression, the use of these techniques has gained significant traction in healthcare - including health systems like the Cleveland Clinic. This session will discuss the use of predictive analytics to go beyond just assessing the risk of a patient, and instead focus on prescribing a specific intervention of care. In practice, the recommendation of a care plan goes beyond just attributing a score to a patient. It provides physicians and care coordinators the most probabilistic options - e.g. referral to social worker, depression screening, weight-loss program - for positively impacting a patient's future healthcare outcome.

Description: Predictive models can improve the traditional actuarial assumption-setting process. Data science techniques can provide deeper insights into experience through their ability to identify key drivers, rank importance and account for interactions. The presenters will cover ways in which companies can utilize predictive analytics to set assumptions, including the various methods available and practical approaches for implementation. They will also discuss the advantages and challenges of using predictive models for setting assumptions for life, annuity and disability products.

Description: The presenters will share information about: -Insights into causes of death through the use of predictive models, based on a reinsured U.S. life portfolio, using logistic regression models to understand three main causes of death: cancer, heart and external causes; -Opioid abuse is a widespread phenomenon with drug-related mortality rates rising steeply in the U.S. over the past two decades. But what does the most recent data tell U.S. about the outlook for Opioid-related mortality?; -'Deaths of despair' have been publicized as a cause for the poor performance of U.S. mortality in recent years. The presenters will be looking at the different causes under this label, their evolution and explore the drivers behind them.

Description: This session will cover clustering techniques generally (k-means, hierarchical), and how k-means have been used to segment policyholders using third party features such as credit scores to assess customer profitability.

Description: This session will review limited fluctuation credibility, and how to apply that approach to a GLM. Alternative credibility approaches through the lens of ASOP 25 will also be discussed.

Description: The presenters will explore policyholder experience data for long-term products, risks of over-fitting data in developing behavioral models, and ways to mitigate these risks and develop a sustainable and coherent framework to translate complex policyholder behavior experience data to behavioral models using company- and industry-level data. At the conclusion of the session, attendees will be able to: - Identify key drivers of policyholder behavior for long-term product - Understand the risks of over-fitting data and a false sense of precision in developing behavioral mode - Understand techniques that can be used to mitigate these risk