Improving Learning Through Online Data Mining Contests

Author Information
Author(s): 
Galit Shmueli
Author(s): 
Rajib Saha
Institution(s) or Organization(s) Where EP Occurred: 
Indian School of Business
Effective Practice Abstract/Summary
Abstract/Summary of Effective Practice: 

"Business Analytics Using Data Mining" is an elective MBA course at the Indian School of Business intended to equip students with knowledge and basic experience about using data mining for tackling problems or opportunities in an organization, as they become managers of "data crunchers". Important course objectives include (1) obtaining hands-on experience of using data mining, (2) understanding the boundaries of "data crunchers" and their managers, and (3) identifying business opportunities for using data mining. Using online data mining contests can greatly assist in achieving these three objectives, in addition to supplying an exciting and rewarding experience for students.

Online data mining contests, such as those run on Kaggle.com, have emerged as an exciting meeting place for industry, practitioners and data science enthusiasts. Typically, a company posts a dataset and an intended goal and set an award and submission deadline. Participants from around the globe can participate and submit entries. By integrating an online contest into the course, students learn about problems of interest to industry, get to try out their data mining skills, and observe professionals tackle the same tasks.

Description of the Effective Practice
Description of the Effective Practice: 

Practice: Integrate an online data mining contest into a data analytics course. This can be done either by taking advantage of an already running contest on a platform such as www.Kaggle.com or www.CrowdANALYTIX.com, or by creating a dedicated contest entry on such a platform and opening it up to students and to the world.

Grading: Student teams are requested to submit at least one entry to the contest before a given deadline. Grading is based on the entry exceeding some baseline score, which verifies that the student team has indeed exerted the needed effort and used the learning to build a model that achieves the contest's task.

Supporting Information for this Effective Practice
Evidence of Effectiveness: 

Interactions with student teams during and after the contest have shown that this active learning component is greatly appreciated, both in terms of excitement and in terms of testing their learning and skills in a realistic situation. An informal course survey has also indicated the same.

Team performance was overall good, and a few even won a contest that had outsiders participating.

Since the course also includes a team project, students in courses that integrated an online contest had more experience and an understanding of the connection between the business domain and the data analytic framework.

How does this practice relate to pillars?: 

Learning effectiveness: An online contest is an active learning activity that integrates many of the topics and skills taught in the course. It is a complex task that requires team work, integrative thinking, technical skills, and communication. The immediate feedback (and possible awards) strengthens learning effectiveness, both short term and long term.

Student satisfaction: The contest builds excitement among students. The competitive nature of it, and especially competing not only with peers but also with professionals around the world, leads to satisfaction and an atmosphere of fun. Highly motivated student teams also engage voluntarily in other contests.

Access: The online contests give students access to two important stakeholders in the business analytics domain. First, they see the types of tasks and problems that companies/organizations post as contests, and therefore can better understand the need of the market. Second, they compete against professional data scientists and other data aficionados, and can compare their skills to the others (by looking at the scoreboard).

Faculty satisfaction: The student excitement is contagious! Students also become more active in reaching out to the instructor in order to improve their performance.

Scale: The contest platforms are built to scale to many participants. Integrating a contest can be done with any data analytics course, even very large classes. By using an existing free and online platform, the scaling issue is immediately supported.

Equipment necessary to implement Effective Practice: 

Internet access and computer with browser.

Estimate the probable costs associated with this practice: 

None, if using an existing contest.
To set up your own classroom contest requires investment of time, but is free of cost.
Optional: It is possible to assign awards beyond those in the contest.

Contact(s) for this Effective Practice
Effective Practice Contact: 
Galit Shmueli
Email this contact: 
galit.shmueli@gmail.com