M3 MathWorks Math Modeling Challenge

Abhay Sri
Feb 28, 2021
2 min read

Yesterday, I had the opportunity to participate in the M3 MathWorks Math Modeling Challenge, and it was a lot of fun. My team consisted of me and four of my close friends, and although while the outcome was unexpected, we learned plenty and are expecting to participate in it next year as well.

Prior to competing in the competition, we all assumed that it would be more mathematically oriented, and our primary objective would be to create or research formulas that would solve the problem presented by the competition. However, as it would turn out, the program heavily involved the use of data collection, refinement, analysis, and modeling. This turned out to be a pleasant surprise and gave me an edge in our team. The problem for this year centered around networking, primarily estimating bandwidth and its costs, as well as the distribution of Wi-Fi.

At first glance, the problems presented seemed simple. However, once we started working on them, they looked almost impossible. The contest provided sample datasets for us to use, however they did not contain nearly enough data. As a result, we were forced to research data sets on bandwidth cost, and we had to make sure they were from the same source as the data provided. For context, the website where we collected data from is here: https://www.newamerica.org/oti/reports/cost-connectivity-2020/. The website was neat, but it only provided data for 2020. Consequently, the data was expansive, it was not sufficient enough to create a model for bandwidth cost over time. As a result, we had to find the cost of connectivity for different years, so we could compare them.

After searching for the much-needed data for a while, I realized that you could change the link to the year desired. While some years like 2018 and 2017 did not work, we acquired the data for 2012, 2013 and 2014. With 4-5 datapoints for each year, we were ready to create a model predicting the cost of connectivity in the future. We used Casio’s exponential model creator and created a model that had a R value of .96. Needless to say, our hard work paid off. However, it came at a cost. By the time we finished question 1, the competition had just hours left. Consequently, we were unable to make considerable progress on any of the other questions.

All in all, this experience was educational and extremely beneficial. It was not only a way for me and my friends to do some group work during COVID, we also better learned how to collect, process, and refine data. I would highly recommend attending the M3 competition, although you will have to wait until next year to do so.