Interesting application of machine learning to understand high MOOC dropout rates
The Data Until I Die! blog had an interesting post a few weeks ago about the KDD Cup competition to understand a very large dataset regarding MOOC course completion rates (~80,000 students' data), and the authors own attempt to compete. The author used machine learning approaches to predict the probability that a given student would drop out... check out the post for a nice narrative of the process and findings... but here are some interesting graphs to pique your interest:
(images from Data Until I Die!)