Predictive Analytics Pilot

Big data is a daily news topic. Actuaries, with their deep understanding of statistics and models designed to assess risk, are primed to be leaders in the application of predictive analytics techniques. Market research and stakeholder discussions reveal predictive analytics is a topic of major interest for post-qualification actuarial education.

As part of a broad set of analytics initiatives, in March 2016 the Society of Actuaries (SOA) began the process of developing a rigorous and specialized pilot certificate program for actuaries in predictive analytics. The pilot program targeted SOA members (FSAs) with some experience in the field. A curriculum mapping process produced a set of learning objectives. The delivery format was set as an online interactive experience with a rigorous project-based assessment. To construct the program in a timely manner, two vendors were selected as the result of a comprehensive request for proposal (RFP) process. Willis Towers Watson provided the subject-matter expert team that created the education content to address each of the learning objectives. Wise Wire, a learning experience design company, transformed that content into an online education experience.

With content development well underway, a call for pilot program participants was issued on Jan. 31, 2017. The SOA received nearly 100 applications. To better manage evaluation feedback and ensure participant interaction, the pilot program was limited to 30 participants. With oversight from a volunteer advisory group, 30 applicants were selected from a variety of employers and practice areas.

Curriculum

The online portion of the course was made up of six modules and was launched on April 10, 2017. The module topics were:

• Module 1: Predictive Analytics Tools
• Module 2: Effective Problem Definition and Project Management
• Module 3: Data Design, Transformation and Visualization
• Module 4: Data Exploration
• Module 5: Feature Generation and Selection
• Module 6: Model Development and Validation

Within each module, there were knowledge checks, exercises, end-of-module tests and opportunities to interact with other participants via a private discussion forum. At times, the module instructions would ask for specific interactions. At any time, participants were free to use the forum to make comments or ask for help.

The participants worked on a variety of data sets, using RStudio to perform the analyses. Sometimes full code was supplied, sometimes partial code needed to be filled in or altered, and other times code needed to be written from scratch.

The concluding learning experience was a two-day seminar. Day One included a recap of the module content, small breakout groups for additional discussion and practice, and numerous Q&A opportunities with the facilitators. The assessment on Day Two was a six-hour project involving the solution of a business problem using a subset of a larger data set familiar to the participants from their module work. A large component of each participant’s final score was based on demonstrating the ability to explain concepts, methods and results.

Grading followed usual SOA processes. Papers near the pass mark were graded by a second person, with the two graders reconciling differences. Results were presented as pass/fail. Those who participated in the seminar and were unsuccessful will have the opportunity to try again at a 2018 program.

Feedback

Our pilot group participants and their employers committed to providing detailed feedback on each module and the program as a whole. Feedback was obtained through a variety of survey tools. Viewpoints were obtained not only on specific content items, but also on whether the pilot group and their designated managerial contacts supported formal implementation and, if implemented, what changes to make for the next version.

All of the feedback was compiled and shared with the SOA’s Professional Development Committee (PDC). The PDC was satisfied that the positive feedback combined with the demonstrated interest in a certificate program, and it formally recommended continuation to the SOA Board. At its October meeting, the Board approved making the certificate program permanent, with at least two offerings in 2018. In addition, the PDC is to research topics for a second certificate program.

For 2018, the program will retain the primary feature of a group of participants progressing synchronously through the e-Learning component and culminating in a two-day seminar that will include a formal assessment. Each program will be designed to host 50 participants. Program registration will be open to all credentialed actuaries. Announcements on program dates and other details will be posted to the certificate program webpage.

The certificate program is a key development in the full suite of offerings in predictive analytics. The SOA is committed to ensuring that actuaries are well-positioned to take a leading role in using these tools in insurance and other risk management applications.

Related Links

Learning Objectives
Predictive Analytics Certificate Program
Predictive Analytics