A Process for Predicting MOOC Attrition

The goal of this shared task was to predict attrition in a MOOC through use of the data and logs generated by the course. Our approach to the task reinforces the idea that the process of gathering and structuring the data is more important (and more time consuming) than the predictive model itself. The result of the analysis was that a subset of 15 different data features did a sufficiently good job at predicting whether or not a student would exhibit any activity in the following week.