Professional Development

Predictive Analytics Seminar – February 27, 2019 – Toronto

Program at a Glance

Wednesday, February 27

07:00–17:00    Registration and Information Desk

07:30–08:30    Buffet Breakfast

08:30–08:35    Opening Remarks

08:35–09:35    Session 1 Opening Plenary

09:40–10:40    Concurrent Sessions

Session 2 Privacy and Bias Considerations of Predictive Models
The insurance industry increasingly uses machine learning models to make decisions that affect people's lives. In order to ensure that the models are fair, models should be audited for bias. The metrics available to evaluate fairness vary by the context in which the model is deployed. In this session, we will walk through privacy considerations, metrics to assess fairness, discuss open-source tools for bias audit use, and finish with a case study discussing a deployed model.

Session 3 Are You behind? (CIA/SOA Predictive Analytics Survey Results)
In this session, we will discuss the results from the SOA-CIA-sponsored survey on analytics practices of the Canadian life and health insurance industry. This is an opportunity to understand what other companies are doing and benchmark your practices with those of your peers. Gain deeper insights about potential applications and their perceived value and effort.

Session 4 Practical Aspects of Predictive Models
The speakers will explore the practical aspects of bringing models to production, maintenance, and management through their entire lifespan. Topics covered include production considerations, versioning, documentation, peer review, and monitoring the model over its lifespan.

10:40–11:00    Networking/Refreshment Break

11:00–12:00    Concurrent Sessions

Session 5 Predictive Modelling Pitfalls. . . and How to Avoid Them
Predicting the future is not easy. In this interactive session, we will discuss four critical pitfalls that can turn any predictive model from useful to useless—and at worst, destructive. He will illustrate each predictive modelling pitfall through a short case study, and will discuss how you can prevent or mitigate the impact of each one.

Session 6 Advances in Cyber Risk Modelling
The cyber risk modelling field is rapidly evolving. In this session, our panel will explain the key drivers of cyber risk, the data and modelling techniques used to calibrate and validate cyber-risk models, and the output and use of probabilistic models. By the end of this session, you will know how cyber risk models can support product development, underwriting, pricing, portfolio optimization, and capital allocation for the cyber line of business.

Session 7 Cross-Selling
As companies look for growth opportunities in a saturated, competitive marketplace, selling additional products to existing customers is a very attractive proposition. The challenges become identifying customers who would benefit from additional products, determining the right product for each customer, and tailoring communication so that the customer is receptive to the offer. This session will explore how to use predictive modelling to address these challenges and drive an optimal cross-selling strategy.

12:15–13:00    Luncheon

13:10–14:10    Concurrent Sessions

Session 8 Experience Studies, Mortality Modelling Using General Linear Models (GLMs)
For many years, actuaries have been privileged with robust experience data to use in setting model assumptions like lapse rates, mortality, and morbidity. Recent increases in computing power and the advent of predictive analytics present an opportunity to enhance the traditional experience study, with the objective of greater accuracy of forecasts and improved granularity of risk segmentation. This session will explore how multivariate statistical analysis and machine learning can improve the experience study process and drive tangible benefits for pricing and valuation. A number of companies have explored different statistical and machine learning techniques for looking at their mortality experience. What have they learned? How did they choose which tools to use and how do they validate this approach versus the traditional methods of the past?

Session 9  Make It Shine: Collecting and Cleaning Your Data
Companies can leverage both internal and external data to support predictive analytics. In this session we will discuss common internal and external data sources that underlie analytics in the insurance industry, as well as techniques to appropriately clean this data prior to analysis.

Session 10 Disability Claims Analytics
Claims scoring techniques for disability claims can be developed by a bucketing process based on likely time to resolution. Claims that can be resolved quickly can be assigned to less-experienced claim handlers, or automatically paid with the goal of saving on expensive resources. Claims that are likely to become permanent should focus on expense management, including the possibility of social security offset. However, it is the claims in the middle that offer the greatest opportunity for management and improvement.

14:15–15:15    Concurrent Sessions

Session 11 Life and Health Underwriting
Many life insurers want to simplify the insurance application and underwriting process to meet the needs of today's client. Limited personal data can be captured from a client and supplemented with third-party data and models to predict other risk factors. Predictive models can also help determine whether an applicant is eligible for a simplified underwriting process. This session will explore how insurers have successfully used predictive analytics to simplify the application and underwriting process while navigating through concerns of adverse selection.

Session 12 And the Winner Is . . . How to Pick a Better Model
You have just finished running data through a predictive modelling package. Now all you need to do is summarize the results, send them along, and you’re done, right? Wrong. At the absolute minimum, you should understand and demonstrate the goodness of fit of the model. In most cases, you should also prove that the constructed model provides lift over the existing rating structure. After all, what good is a new model if it cannot outperform the competition? In this session, we will explore in significant detail three often overlooked components of the modelling process: measuring goodness of fit, assessing lift, and internally validating a predictive model. Key topics include confusion matrices, receiver operation characteristic (ROC) curves, Gini indices, lift charts, double-lift charts, residual plots, likelihood, penalized likelihood, and deviance. Model development is usually a major investment. We should make sure our models perform well to get the best bang for the buck.

Dangers of overfitting: Simple mathematical equations outperform very complex ones in predictive analytics situations. This session will also cover why and how to guard against that and other items that might not be intuitive.

Session 13 Fraud Detection
The speakers will describe how to use predictive modelling techniques to detect fraud. In this session we will nuance how predictive fraud analytics differs from and can be built upon traditional approaches and will provide actionable insights on how to enhance your fraud detection processes with these techniques.

15:15–15:30    Networking/Refreshment Break

15:30–16:30    Session 14 • Plenary Session: Keynote Speaker David Coletto

There are powerful new forces at work in the modern workplace and consumer market, a dynamic driven by the intersection of disruptive technology and a new generation of consumers and employees who were raised differently, have different values, and different expectations.

This revolutionary, generational change represents enormous risk but also creates an incredible opportunity for the insurance industry and actuaries.

As one of Canada's foremost experts on generational change and youth, Mr. Coletto has spent over eight years as the founder and CEO of Abacus Data trying to understand his generation and working with brands, associations, and public sector organizations to reorient themselves for this millennial-dominated world. He is convinced that a generational analysis is critical to understanding the forces at play in the workforce.

As Canada's leading authority on his generation, Mr. Coletto promises to deliver an engaging and data-filled presentation that explains why using his “SHIFT” lens  is critical to leading and succeeding in today's marketplace. 

Don't miss the founder of Canada's only research firm dedicated to helping organizations navigate the unprecedented threats being caused by the generational change in Canada and around the world.

16:30–17:30   Networking Reception

Sponsorship Opportunities

To learn more about sponsorship opportunities, please contact Kelly Fry.