DETECT: A Hierarchical Clustering Algorithm for Behavioural Trends in Temporal Educational Data
Project Overview
The document discusses the innovative use of generative AI in education, particularly through the introduction of DETECT, a hierarchical clustering algorithm that enhances the analysis of educational data by focusing on temporal aspects to uncover behavioral trends among students. This advanced algorithm addresses limitations of traditional clustering approaches, enabling educators to identify patterns in student engagement and challenges over time. Through case studies in programming courses, the paper illustrates how DETECT can provide valuable insights that lead to improved educational outcomes, thereby highlighting the potential of generative AI to transform educational practices. The findings suggest that utilizing such AI-driven tools can enhance understanding of student behaviors, ultimately fostering better support and tailored learning experiences in educational settings.
Key Applications
DETECT
Context: Online programming courses for students (N > 600), including beginner programming courses (N = 635), focusing on student behavior during programming exercises, particularly identifying trends and challenges related to student engagement and dropout rates.
Implementation: Applied DETECT to analyze and cluster time-series data of student behavior during programming exercises, including the identification of specific exercises that correlate with disengagement and dropout rates. The approach involved determining key behavioral patterns, such as autosaves and completion rates, to provide insights into student engagement.
Outcomes: Identified trends in student behavior over time and highlighted specific exercises that were particularly challenging, correlating these with student dropout rates. This provided educators with actionable insights to improve course design and student support.
Challenges: The complexity of student behavior and the diversity of data can make clustering challenging, along with the need to determine appropriate features and objective functions to accurately capture student behavior.
Implementation Barriers
Technical Barrier
The complexity and diversity of student behavior data can complicate the clustering process.
Proposed Solutions: Utilizing flexible and customizable objective functions in DETECT to better capture relevant trends.
Data Quality Barrier
Incomplete or inconsistent data can hinder effective clustering.
Proposed Solutions: Implementing strategies to handle missing data and ensuring robust feature selection.
Project Team
Jessica McBroom
Researcher
Kalina Yacef
Researcher
Irena Koprinska
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Jessica McBroom, Kalina Yacef, Irena Koprinska
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai