Detection of Online Learners at Risk of Attrition

With the increasing popularity of online learning, it can be challenging to keep and retain students. It is, therefore, increasingly important to identify students who are at risk of attrition early in order to keep them engaged in courses and to support them in achieving their learning goals. The client for this project, the Microsoft Worldwide Learning organization, has estimated that it is more expensive to acquire a learner who has dropped out of a course than to initially attract them. Once learners leave a course, they do not normally come back. The organization has asked to study if it is possible to predict when learners are likely to drop out of Microsoft’s massive open online courses (MOOC) before they do. Being able to identify learners who are at risk will allow the teaching team to undertake a range of nurture actions that are student-focused. Actionable insights about the company’s various courses will also allow management to improve their decision making in handling learner churn and retention. Using data from one of the company’s more popular MOOC courses, this project created multiple prediction models to identify students with a high probability of attrition. Evaluation of these models was performed in order to select the best one. Additional metrics for monitoring the success of the course were included in a dashboard to create actionable next steps for the client. The results from this client project can be implemented to scale to other courses to improve the learning experiences for their students.