Negative link prediction to reduce dropout in Massive Open Online Courses

Area of Science:

• Educational Technology
• Computer Science
• Data Mining

Background:

• Massive Open Online Courses (MOOCs) face challenges with high student dropout rates.
• Existing dropout prediction methods often rely on extensive student activity data and supervised machine learning.
• Graph-based algorithms show promise in identifying relationships with limited data.

Purpose of the Study:

• To propose a novel, low-complexity negative link prediction algorithm for early dropout prediction in MOOCs.
• To utilize solely network topological data (enrollment information) for predicting student attrition.
• To evaluate the effectiveness of graph-based approaches for MOOC dropout prediction.

Main Methods:

• Data was converted into a graph, representing students and courses as nodes and enrollments as links.
• A novel negative link prediction algorithm, leveraging network topology, was developed.
• The proposed method was compared against baseline approaches using experimental data.

Main Results:

• The proposed graph-based method achieved significant performance improvements over existing baseline methods.
• The approach demonstrated competitive and promising results when tested with a supervised link prediction framework.
• The study confirmed the utility of network topological data for dropout prediction.

Conclusions:

• The novel graph-based negative link prediction algorithm offers an effective and efficient solution for MOOC dropout prediction.
• Utilizing only enrollment data provides a viable alternative to complex methods requiring extensive activity logs.
• Future research directions are identified to further enhance prediction accuracy and model robustness.