Join us on Monday, February 19 from 12:00 to 1:30 p.m. in Space 2435 of North Quad (105 S. State St.) for the first gathering of AIM Analytics: MOOC Dropout Prediction Challenge.
In the meeting, we will talk about the data, how to access to MOOC Replication Framework (MORF), and some demo if time permits.
If you are interested in accessing MOOC Dropout Prediction Challenge, please sign up using THIS FORM and let us know if you can join us in the meeting on the 19th. Even if you cannot join the meeting, please sign up so we can keep you updated on the prediction challenge.
Academic Innovation at Michigan Analytics workshop series (AIM Analytics) is a bi-weekly seminar series for researchers across U-M who are interested in learning analytics. The field of learning analytics is a multi and interdisciplinary field that brings together researchers from education, learning sciences, computational sciences and statistics, and all discipline-specific forms of educational inquiry.
MOOC Dropout Prediction Challenge
The AIM Analytics prediction challenge is focused on tapping into the wealth of data science talent at the University of Michigan in the service of building the best dropout prediction modeling approach. The challenge is open at all students, staff, or faculty at the University of Michigan (undergraduate through graduate), and runs through April 23rd, 2018, with a $500 prize provided to the team which provides the best model. In addition, the AIM Analytics bi-weekly seminar series will host a number of collaborative events aimed at explaining data, approaches already taken in the field, and technical advice on how the problem might be solved.
The University of Michigan has been a leader among universities in the offering of MOOCs, and currently has more than 90 different MOOC initiatives on topics including technical (programming, data science), the humanities (literature), social phenomena (negotiation, business), and medical (oncology). As part of this activity, the University of Michigan has collected data from different courses run on the Coursera platform. These courses are made up of real learner interactions and are provided by the vendor. Each course has its own pedagogical approach, assignment criteria, and set of learning resources. A description of the dataset schema is available courtesy of the University of Illinois, Urbana-Champaign.
A single course offering will be provided to participants under a data use agreement to allow for exploration of data.
For this task, dropout will be considered as a student who completes the third week of the course but does not show any clickstream activity in the final week of the course. Labels are provided for all instances for training and testing.
Note: Participants are given access to full course data for training. By the end of February, an evaluation script will be provided which can be used to truncate data to three weeks (which is what the holdout set has)
Models created by participants will be run on the MOOC Replication Framework (MORF), created through a collaboration between the University of Michigan School of Information and the University of Pennsylvania Center for Learning Analytics.
This framework allows participants to package their analysis software and run it on a hosted infrastructure where data is held embargoed from the participant. Results are returned using a variety of metrics (AUC, F1, precision, etc.). A hold out set of one session offering for each of the courses in the dataset will be used for the final evaluation with median AUC across offerings being used for the final evaluation. Like Kaggle, we will provide a public leaderboard of evaluations (in development). Unlike Kaggle competitions, participants are unable to see their final predictive performance until the competition has ended.
To register for this event, please sign up using THIS FORM.
Lunch will be included.