Hack-A-Thon
This article will share the history of the SOA’s Predictive Analytics and Futurism (PAF), Modeling, and Actuary of the Future Sections’ sponsored Hack-A-Thon (HAT) and how it has evolved over the years. The article describes the problems from past competitions, their objectives, and this year’s results. Lastly, we will discuss how it works behind the scenes for people that may want to volunteer in planning the next HAT.
What is a HAT?
This year we had a great predictive analytics HAT in October. The HAT is a live four-hour predictive analytics competition sponsored by the Society of Actuaries. This year, we awarded 10 prizes (US$50 Amazon gift cards).
How Did it Start?
The HAT started because of a 2018 PAF survey on what its members wanted. Two major conclusions were drawn from the survey: Members wanted to learn predictive analytics and meet other people. From this outcome, PAF thought it would be fun to do a HAT. The HAT has grown every year. The beginnings are humble, but persistency is paying off. The first two years saw around 18 to 20 people, but this year we had 40. We were very happy with the turn out. We are hoping to continue growing from here.
When creating the competition, we try to do something that is related to actuaries, but not so closely that it is likely someone would be exposed to it at work and that competitor would blow the competition out of the water. Every year we allow people to work separately or in groups of four or less. You can join with your friends or be randomly assigned to a team.
What is the Most Enjoyable Part of HAT?
My favorite part of the competition is pulling all competitors together at the end. I always enjoy getting interactive feedback from everyone. The interaction is always fun and lively. We ask, what they liked about the competition? Was the challenge appropriate level of difficulty? What models and languages did people use? Every year we have worked hard to incorporate the ideas to make the next year even better.
Tell Me About the 2019 Competition!
The 2019 competition used the Public Use Microdata Sample (PUMS) files from Census.Gov. This demographic data was used to try to predict which people from the population were most likely to be actuaries. For this competition, we asked everyone to assess their skill level. We made sure to place the more experienced with the less experienced, so beginners could learn from the more experienced people. The most experienced team in the room was trying to use a random forest for the competition. The winning team of the competition used plain old actuarial judgement and Excel. The demographic data was the highest degree of attained education and area of study. They guessed that anyone with at least a bachelor’s degree and in a scientific field was likely to be an actuary. This was not the outcome that the organizers expected, but we got a good chuckle at it none the less.
At the beginning of the 2019 competition, we had an hour-long presentation because the event organizers needed time to set up the competition room. I noticed that there was some anxiety in the room, so as an ice breaker I asked everyone where they worked, why they signed up for the competition, and their perceived level of experience in predictive analytics. The answers that most everyone gave was that they were beginners that wanted to learn from each other. Once everyone realized they were in the same boat, the anxiety immediately lifted from the room, and everyone had fun.
Tell Me About the 2020 Competition!
Given that almost everyone was a beginner in the 2019 competition, we decided to tailor the 2020 competition to beginners. This was the first year that we coordinated the HAT with the Practical Predictive Analytics Seminar (PPAS). We thought this was important because it would be a fun way to solidify the information during the course and work together with other peers. We made sure that the competition was attainable from course material. The 2020 competition used SOA student performance data. The goal was to use demographic data to determine Associate of Society of Actuary (ASA) attainment.
Tell Me About the 2021 Competition!
For the 2021 HAT, we had a major change in the format. We broke the competition into beginner and expert groups, instead of doing a one size fits all. There is every intention to continue this into the future. We do attribute this change to a better turnout, because beginners will be less intimidated to sign up, and experts can be challenged too. We again tied the beginner competition to the PPAS. (Participants do not have to attend the PPAS to do the beginner competition. It is strictly to help reinforce the material from PPAS.)
The beginner competition used bank data from Kaggle. They tried to predict whether customers would stop using the banking product based upon demographic, social, and behavioral features.
The expert competition used Bureau of Labor Statistics American Time Use Data: 2018–2020. They predicted, based on how people spend their time, whether the study participants are in business and financial operations or computer science and mathematics. (There was only one actuary in the dataset, so we tried to approximate with a group from larger data.)
The most memorable outcome from this competition is that a mistake was made to check the correlation of all the data with the response variable in the test set. One of the columns was 100 percent correlated with the response column. Three teams won the expert competition, because they exploited this issue.
Volunteers Needed!
Every year around March, we start planning for the next HAT. Every year it takes a team of volunteers to bring the HAT to fruition. We take what we learned from the previous year and always try to improve it. It is so fun to work together and meet new people. Even if you don’t feel confident in predictive analytics, you can still get involved. It is great to have different perspectives helping make the competition. Every year for the beginners we try to have one or two experts to answer questions if they need someone to bounce ideas off. In 2021 we were not able to find any experts for this role, so participants missed this feature of the competition. If you have a desire to volunteer, you can contact Agnes Niemas at aniemas@soa.org or Jane Lesch at jlesch@soa.org.
Conclusion
Now that you have learned about the HAT and its history. I hope this piques your interest to sign up for the competition this year. The feedback has always been very positive that people had fun participating in the competition. We are always looking for volunteers. I hope to see you this year.
Statements of fact and opinions expressed herein are those of the individual author and are not necessarily those of the Society of Actuaries, the newsletter editors, or the respective author’s employer.