UCL Data Science Student Challenge
Saturday 11th to Sunday 12th of March, 2017
Guest blog by Rohan Kopparapu (4th year UCL) and Microsoft Student Partner
Microsoft’s global series of data science hackathons made its second stop in the UK, at UCL last weekend. http://ucl.dsschack.com
Over 170 students from across the country joined us, despite TfL closures, at the Bloomsbury campus on Saturday, to take part in the 21-hour long challenge.
Students formed teams up to a maximum of 4 people to develop a data science solution against a dataset and challenge provided see http://aka.ms/ucldssc
The hackathon was themed around the development of innovative data science solutions with large financial data-sets in order to give Bloomberg a competitive advantage in the finance industry, while demonstrating use of Microsoft’s Cortana Intelligence Suite. Students had the opportunity to work on deep learning, text-mining and NLP to extract information from news articles in real-time.
Laura da Silva giving a demo on the basics of Data Science with Azure ML in the Gustave Tuck LT
The challenges also focussed on the use of machine learning and classification of news articles for ranking and sentiment trend analysis, while demonstrating the impact of news and events on asset prices, as well as understanding the trends between various instruments in financial markets.
Hacking underway in the Wilkins Building Cloisters
Teams were judged on their ability to identify significant relationships between instruments, news reports and trends, as well as illustrating the strength of those relationships. How that information could be used to enhance decision making, and identify generalised patterns of events were also important considerations.
The winning team, HypeTrain – Utku Ozbaluk, Alex Young and Lukas Weiss, Masters students in Data Science from the University of Southampton, came up with a forward indicator, which isn’t currently offered by Bloomberg, more information can be seen here.
Bing is King placed second in the challenge – Brian Gunawan, Jason In, Raymond Tan and Wayne Tsui, 2nd year undergraduate Computer Science students from UCL, came up with an engine that answers questions using predictions with boosted decision tree regressions, using additional data from Reddit, see here.
Third placed Kernel Trick – Jaromir Latal, Faiz Abdul, Anirudh Pillai and Mujavid Bukhari, also 2nd year undergraduate Computer Science students from UCL, demonstrated the effect of news and events on asset prices by building a neural network to predict the return of a stock by looking at relevant news sentiments for the current day, see here.
From top à bottom, left à right: judging, 3rd place, 2nd place, winners
All the Hack entries are at http://www.github.com/DSSCHack2017
Thanks to the team
MVPs - Richard Conway (Elastacloud) and Jen Stirrup (Data Relish), Laura da Silva (Elastacloud), Paul Gaskell (Bloomberg), Terry Christiani (MSFT Redmond), MSFT UK Academic Relations - Geoff Hughes and Lee Stott mentored the teams, helped polish their hacks, and then acted as judges for the event.
UCL Computer Science students, Rohan Kopparapu (4th year), Jasmine Lu (3rd year), Vika Christy (3rd year), Christopher Hammond (2nd year) and Shanice Ong (2nd year) led a dream team of students to design, resource and implement this fantastic event, supported by Vumero Institute.
The team behind the hackathon
“As an employer of data scientists, I was overwhelmed by the solutions of many of the teams. The level of competence and creativity in the solutions that were presented back to me was fantastic. UCL is clearly fostering the academic benchmark for the next generation of data scientists. It was fantastic to see the depth of collaboration between different academic disciplines, from economists to statisticians to enhance the flavour of their solutions. Many of the students managed to grok some complicated Microsoft machine learning tools in a short space of time, showing the levels of agility necessary for real world data science careers”, said Richard Conway, http://www.Elastacloud.com
- 29% of attendees were from a non-CS background
- Over 50% of attendees were in their first or second year of university
- 38% of the attendees had never been to a hackathon before
- 90% of attendees were fairly well versed in programming
- Over 700 slices of pizza, 450 pies, 200 noodle pots and many, many snacks were consumed
- The Facebook event page reached over 21k people and over 4k unique views in just a week https://www.facebook.com/dsschack