Eight Teams Compete in HSE’s First Data Science Hackathon
HSE University has recently hosted an SAS Data Hack Platypus data analysis hackathon, where teams competed to solve a case problem provided by the online hypermarket, Utkonos (or, in English, ‘Platypus’).The top five finalists received devices from partner companies as gifts.
The hackathon was organized by the HSE Faculty of Computer Science together with SAS and the Utkonos online hypermarket. Within 23 hours, the final teams of the qualifying stage had to use behavior data from the store website in order to predict how long it takes a customer to change his or her mind and cancel an order.
The task of the final round was in many ways a more complicated version of the qualifying round’s task. In the final round, teams had to perform data analysis in order to predict the probability of whether a customer would cancel his or her order in the first place. ‘Cancelled orders are a big problem for online retailers,’ says Danila Naumov, Director of the Utkonos company. ‘For a long time Utkonos was the only online grocery delivery service. The emergence of a large number of competitors a couple years ago prompted us to work on digitally transforming our company in order to improve our services and optimize internal processes.’
If a customer cancels their order after it has been processed and prepared for delivery, the company incurs financial losses. In addition, the items are blocked to other users, so there is also the so-called lost demand. What’s more, perishable food items that were in the warehouse and then went out for delivery and driven around the city for nothing have to then be written off or sold at a discount.
Danila Naumov
It was difficult for us to choose a task for the hackathon, since most of our practical cases require an understanding of the specifics of the business. Creating a model isn’t enough—we need something that can be put into use. In order to come up with a different way of doing things for a business, one needs to have special knowledge. In the task we designed for the hackathon, there was more opportunity for creativity and putting one’s data science skills into direct use.
In the final round, the accuracy of the resulting models played a large role; we ranked the teams according the number of errors they made. In addition, in the presentation stage, teams could earn up to 2 points for the model’s applicability and the quality of their presentation. In my opinion, 50% of success, and sometimes more, depends on emotional intelligence. It is necessary to be able to communicate with contractors and to successfully pitch your idea to the business.
First place in the hackathon went to a developer from Norilsk, Dmitry Fedotov. He found that the biggest sign of order cancellation is the presence of similar orders for the same delivery date in the user’s account.
Dmitry was able to most accurately interpret most of the data. ‘I have been participating in hackathons for half a year now: I’ve done 15 online competitions and three offline. I don’t know anyone from Moscow, so I’m a team of one. The final round coincided with some vacation time, so I decided to travel to Moscow for it, and I do not regret it – the task was really interesting and the competition was well organized,’ says Dmitry.
Ilya Gorshkov, a master’s student in ‘Data Analysis in Biology and Medicine’ at HSE, took part in the hackathon for the first time. He was also a team of one and took second place. Third place went to two teams of adult developers, and fourth and fifth places went to student teams from Moscow State University.
According to Elena Schenyavskaya, a representative of SAS Russia / CIS, she and her colleagues were interested in hearing the participants’ new ideas and telling them about their internship programme. ‘Hackathons allow you to gain experience solving problems, develop your own style, and evaluate how a particular model manifests itself in practice,’ says Elena Schenyavskaya. ‘In fact, any of the hackathon participants could test their hypotheses with two clicks in our programme, which is also available to students, but they preferred using Python. SAS provides business software that helps people without programming skills develop complex models. Thus, even business analysts without special knowledge have the opportunity to get a decent result using the data mining model.’
After the award ceremony, team members who had advanced to the final round received job and internship opportunities from the organizing companies.