Proactive Student Persistence Prediction in MOOCs via Multi-Domain Adversarial Learning

EasyChair Preprint 7024

Abstract

Automatic evaluation of a student’s STEM learning profile to understand her persistence is of national interest. In this paper, we propose an early ``dropout" prediction model that can identify the potentially `marginalized' student learning patterns to facilitate early instructional intervention in Massive Open Online Courses (MOOC) learning platform. Note that in the MOOC setting, building a comprehensive learning profile of the students is particularly more challenging due to the lack of available information and constrained communication modes. Unlike most existing works, which ignore these environmental constraints of missing information to formulate an over-simplified problem of `one-time' prediction task in a supervised setting, the proposed model introduces a continual automated monitoring and proactive estimation process, which transforms its decision making capacity over time with evolving data patterns. In a semi-supervised scenario, the Multi-Domain Adversarial Feature Representation (mDAFR) strategy promotes the emergence of features, which are discriminative for the main learning task, while remaining largely invariant to the data sources (course from which the data was captured) in consideration. This ensures an enhanced distributed learning capacity over different course environments. Compared to transfer learning, mDAFR reports 11-15% improved classification accuracy in KDDCup dataset, and demonstrates a competitive performance against several state-of-the-art methods in both KDDCup and MOOCDropout datasets.

Keyphrases: Adversarial Learning, Classification, Domain Adaptation, KDDCUP dataset, MOOC, Multi-feature learning, Transfer Learning, adversarial feature representation, course specific learning activity, feature representation, learning activity, multi domain adversarial, student activity detail, target domain

@booklet{EasyChair:7024,
  author    = {Sreyasee Das Bhattacharjee and Junsong Yuan},
  title     = {Proactive Student Persistence Prediction in MOOCs via Multi-Domain Adversarial Learning},
  howpublished = {EasyChair Preprint 7024},
  year      = {EasyChair, 2021}}