I was in two minds about entering the SAS Cortex Challenge. It has been a while since I did the Machine Learning subject in my Master of Data Science course. As it was held during the O-Week of the next subject in my course I felt that I had the time to give it a go. The cost was USD50 (discounted from USD100) for 20 hours of access to the virtual learning environment.

The game scenario was working on a fundraising campaign for a 12-year old, not-for-profit charitable foundation with a million members. The foundation had decided to add a direct contact campaign to its list of marketing activities. Using the predictive modelling software, SAS Enterprise Miner,  I had to predict how many and which individuals to target in the campaign. The objective was to fundraise the most in donation amount given the costs of calling members (sum of predicted amount given minus costs).

Each day there was a webinar hosted by SAS. Days 1, 2, and 4 had tips on how to play the game with days 3 and 5 for Q&A. Unfortunately, due to COVID-19 and everyone seeming to be working from home, my internet quality/speed dropped dramatically and I was unable to attend the webinars live and had to rely on the recordings sent out a couple of hours later.

Whilst I didn’t win the AUD1000 first prize I did learn a little about machine learning in Enterprise Miner. And this was the real reason for doing the Challenge. Enterprise Miner is quite easy to use. The difficulty, of course, came in how to tune the hyperparameters, and the bulk of the 12 hours I used was spent waiting for the jobs to run. I also had problems with the internet dropping out but luckily on reconnection I hadn't lost any work. (In fact, it was so bad one day I had to hotspot my Mac to my phone.)

I managed to increase the operating surplus each day and finished at 6 pm on day 5 in fifth position on the leaderboard.