A new competitive data science student team has been formed at Michigan to solve data prediction challenges in competitive venues. The Michigan Data Science Team (MDST) is one of the first collegiate teams of its kind, with a mission to compete against professional and amateur data scientists from around the world in online prediction challenges.
Every day, vast amounts of data – 2.5 quintillion bytes per day, according to IBM – are generated. Data science techniques are being developed and leveraged to analyze and utilize that data in order to make otherwise unseen connections that can lead to accurate and useful predictions.
At the same time, competitive data science has become increasingly prominent with the immense popularity of high-profile competitions such as The Netflix Prize. Now, online venues such as Kaggle, DrivenData, Quantopian, and others provide platforms for data scientists around the world from which to make impactful contributions to a variety of prediction problems while competing for cash prizes. These competitions have explored prediction problems in healthcare, particle physics, finance, and countless other domains, and have involved many types of structured and unstructured data.
MDST formed in Fall 2015 and has been attracting dedicated students who are interested in taking part in data science competitions, according to Jonathan Stroud, CSE graduate student and MDST Organizational Chair. According to Stroud, engaged students have tended to posses some background in computer science, mathematics, and statistics, and all students with an interest are welcome to participate. MDST meets weekly to share strategies and give tutorials in preparation for challenges. The group offers internal prizes to members who achieve the highest performance in prediction challenges.
Excellent Placement in First Competition
For its first competition, MDST fielded teams in the Springleaf Marketing Response Challenge, which concluded on October 19. 2,225 teams from around the world competed. The top three performing MDST teams and their placements in the competition were:
#34 - Cantseetherandomforestforthetrees: Jared Webb and Alexander Zaitzeff
#545 - Physteam: Arya Farahi and Anthony Kremin
#552 - GGBrown: Xiang Li, Xinyu Tan, Jianming Sang, and Tianpei Xie
MDST is advised by Prof. Jacob Abernethy, whose research interests are in the areas of machine learning, game theory, decision theory, optimization, market mechanism design, and financial applications. He is particularly interested in how algorithms utilized in ML, such as those for discovering patterns in data, are strongly related to methods used in large-scale optimization. In 2005, Prof. Abernethy received an NSF CAREER Award for his big data driven research into the relationship between machine learning and microeconomic theory.
Posted: October 23, 2015