WISEngineering: Achieving Scalability and Extensibility in Massive Online Learning

摘要

Massive Open Online Courses (MOOCs) have raised many unique challenges to online learning platforms. For example, the low teacher-student ratio in MOOCs often means lack of feedback to students and poor learning experiences. We present WISEngineering, a MOOCs platform that provides a rich set of features for overcoming these challenges. The system embraces social media for fostering student reflection. Its automated grading system adopts an open-architecture and uses stack generalization to blend multiple machine learning algorithms. A Zookeeper based computing cluster runs behind auto-grading and provides instant feedback. A behavior tracking system collects user behavior and can be later used for learning outcome analysis. We report the design and implementation details of WISEngineering, and present the design decisions that allow the system to achieve performance, scalability and extensibility in massive online learning.