A team consisting of two fourth year Computer Science students and one third year BDat Science student won first place in the annual Data School Hackathon, which took place in collaboration with Standard Bank Lab and Mobylaz in August this year.
Wicus van der Linden and Daniel van Zyl, both BSc Honours students in computer science, teamed up with Christiaan Hildebrand, a third year student in BDatScience, after they had first met each other as undergraduates in Majuba's Men's Residence in 2021.
Wicus says they were interested in taking part in a hackathon challenge oriented around data science, as they have not had as much exposure to machine learning and data science during their undergraduate studies.
“We knew each other well from time spent together in residence, and we were excited for the chance to prove ourselves in a more real-world orientated challenge," Wicus says.
This year the challenge posed to students was to investigate and analyse a set of telemetry data of various taxi vehicles insured by Mobalyz, aiming to gain insight into various characteristics such as driver behaviour, risk, or performance. The data included more than three million events, from ten different vehicles, with observations including g-forces, speed, and location, Wicus explains.
Teams were given ten days to submit their findings. From the more than 40 submissions from 25 teams, six teams were selected for a final round of presentations. Monetary prizes were available for each of the top three teams, as well as a bonus prize for the best performing women's team.
The winning team produced a decision-tree-based risk assessment model, classifying taxi driving behaviour based on similarity to known insurance claim information: “We made use of external datasets for features such as weather conditions and traffic information from Uber. We also analysed driver routes and parking behaviour to identify which vehicles had significant changes in their daily or weekly patterns, possibly indicating driver changes" Wicus explains.
According to Daniel, the hackathon exceeded all their expectations: “To be honest, it was fascinating to have the creative freedom on an open-ended task, and to have data to work with that had real-world implications and stakeholders. I think it really showed us how we as data scientists can get to work with different industries, people and datasets. And that even students can add value to a business market," he added.
The winning team also thanked the efforts made by the School for Data Science and Computational Thinking, Standard Bank and Mobalyz to put together such an interesting challenge.
On the photo above: The winning team, from the left, Wicus van der Linden, Daniel van Zyl and Christiaan Hildebrand. Photo supplied