A team of Sol Plaatje University (SPU) Data Science students has placed among the top 14% of data scientists taking part in a global competition to use historical product and customer meta data to develop a personalised customer recommendation system for international fashion group H&M.
The five students, under the guidance of SPU Data Science lecturer Ms Nontokozo Mpofu, took part in the competition via Kaggle, a data science platform which, among other things, challenges data scientists of all levels (from beginners to experts) to participate in real-life data science competitions.
The competing students were Ngonidzashe Tinago(Male), Risima Da Gama(Female), Hiteko Kevin Maluleke(Male), Rejoice Chitengu(Female) and Skhumbuso Maleka(Male) (all BSc Data Science 2nd Year students).
“We had to compete with data scientists of all skill levels, from amateurs to experts while adhering to due dates of required submissions, with guidance from Ms Mpofu,” Tinago said. “There were 2 952 teams and 3 759 competitors, with a total cash prize of $50 000 (R855 000) at stake.” Despite only being able to take part in the last 1 ½ months of the three-month competition while attending all their Semester lectures full time, the team impressed their lecturer when they finished in the top 14% of competitors.
“Kaggle provided us with four data sets and one sample submission data set, including images of almost each product in the H&M stores, and transactions of each customer for a given period,” Tinago explained. “Since the datasets provided were too big, we had to code on the Kaggle platform, which was not much trouble as Kaggle had all the necessary resources and packages already installed. Kaggle allowed five submissions a day, so from the time we joined we could have submitted a total of 90+. We unfortunately managed to submit only nine in total due to time constraints.”
Chitengu and Da Gama explained that the team’s strategy was to first understand how the platform works, and get to know more about it and what resources it offers to make things easier. “We looked at previous competitions and how previous winners went about their work. We allocated enough time to the project and set deadlines for each task we allocated ourselves. There were days we would meet up and have fun playing for hours to refresh our minds before getting back to work, and that helped us think clearly all the time.”
Maleka said the team learned a lot from competing on an international level. “It showed us that coding or programming can never be a one-person job – you always need to consult with someone, and you always need to merge ideas for you to get a great outcome. We also learnt that there’s no irrelevant skill in a team, whether it is programming in a legacy language or putting a smile on one’s face, each skill has great significance. We learnt that programming doesn’t necessarily have to be intimidating, it can be fun, but there is no fun if you do not fully understand your data, hence the need for data exploration and analysis.”
Mpofu congratulated her students on their achievement. “Although they didn’t have as much time as the other competitors, they did very good work, and they showed the potential among SPU’s Data Science students. We look forward to rising up the ranks even more impressively in other Kaggle competitions!”