The Third International Workshop on
Graph-Based Educational Data Mining
June 25 - June 28, 2017
*** Title and abstract due May 2nd, 2017 ***
*** Final submissions due May 8th, 2017. ***
- GEDM 2017 proceedings (link)
- Using Graph-based Modeling to explore changes in students’ affective states during exploratory learning tasks (link)
- A Social Network Analysis on Blended Courses (link)
- Mining Frequent Learning Pathways from a Large Educational Dataset (link)
- Learning From Your (Markov) Neighbor: Network-Based Grade Prediction in Undergraduate Courses (link)
Graph-based data mining and educational data analysis based on graphical models have become emerging disciplines in EDM. Large-scale graph data, such as social network data, complex user-system interaction logs, student-produced graphical representations, and conceptual hierarchies, carries multiple levels of pedagogical information. Exploring such data can help to answer a range of critical questions such as:
For social network data from MOOCs, online forums, and user-system interaction logs:
- What social networks can foster or hinder learning?
- Do users of online learning tools behave as we expect them to?
- How does the interaction graph evolve over time?
- What data we can use to define relationship graphs?
- What path(s) do high-performing students take through online materials?
- What is the impact of teacher-interaction on students’ observed behavior?
- Can we identify students who are particularly helpful in a course?
For computer-aided learning (writing, programming, etc.)
- What substructures are commonly found in student-produced diagrams?
- Can we use prior student data to identify students’ solution plan, if any?
- Can we automatically induce empirically-valid graph rules from prior student data and use induced graph rules to support automated grading systems?
Graphical model techniques, such as Bayesian Network, Markov Random Field, and Conditional Random Field, have been widely used in EDM for student modeling, decision making, and knowledge tracing. Utilizing these approaches can help to :
- Learn students’ behavioral patterns.
- Predict students’ behaviors and learning outcomes.
- Induce pedagogical strategies for computer-aided learning systems.
- Identify the difficult level of knowledge components in the intelligent tutoring systems.
Researches related to questions can help us to better understand students’ learning status, and improve the teaching effectiveness and student learning. Our goal in this workshop is to bring together researchers with special interest in graph-based data analysis to 1) discuss state of the art tools and technologies, 2) identify common problems and challenges, and 3) foster a community of researchers for further collaboration. We will consider the submission of full and short papers as well as posters and demonstrations covering a range of graphics topics that include, but are not limited to:
- Social network data
- Graphical solution representations
- Graphical behavior models
- Graph-based log analysis
- Large network datasets
- Novel graph-based machine learning methods
- Novel graph analysis techniques
- Relevant analytical tools and standard problems
- Issues with graph models
- Tools and technologies for graph grammar (pattern) recognition
- Tools and technologies for automatic concept hierarchy extraction
- Computer-aided learning system development involved with graphical representations
- Use of graphical models in educational data
We particularly welcome submissions of in-progress work both from students and researchers with problems who are seeking appropriate analytical tools, and developers of graph analysis tools who are seeking new challenges.
Sincerely: Dr. Collin Lynch, Dr. Tiffany Barnes, Linting Xue & Niki Gitinabard
We will invite submissions of full papers which describe mature work. We will also accept short papers describing in-progress work or student projects, and poster/demo submissions for those presenting available data, tools, and methods. This last category is particularly targeted at researchers who have data or methods available and are seeking to identify potential collaborators.
Long papers should be from describe mature work and be from 7-8 pages long.
Short papers should be 5 pages long and describe planned research; existing datasets that would be amenable to graph analysis; or mining techniques that are available for wider use.
Poster and demo submissions should be 3 pages and should present an overview of the planned content. Posters and demos should be focused on available datasets and tools and be designed to foster discussion.
Papers should be submitted in the EDM 2017 latex and word formats: http://educationaldatamining.org/EDM2017/submission/
We will organize this workshop as a half-day mini-conference with time set aside for paper presentations, large-group discussion, and individual networking. We will open the workshop with a summary of prior meetings. We will spend the workshop on presentations with a short discussion session.
- 1-1:30 Introduction
- 1:30-2 A Social Network Analysis on Blended Courses
- 2-2:30 Predictive Models with the Coenrollment Graph: Network-Based Grade Prediction in Undergraduate Courses
- 2:30-3 Break and Networking
- 3-3:30 Using Graph-based Modelling to explore changes in students' affective states during exploratory learning tasks
- 3:30-4 Mining Frequent Learning Pathways from a Large Educational Dataset
- 4-4:30 Break and Networking
- 25 April 2017 - Abstract due. (Extended to 2 May 2017)
- 1 May 2017 - Final paper due. (Extended to 8 May 2017)
- 15 May 2017 - Notification of acceptance. (Extended to 22 May 2017)
Dr. Collin Lynch
Assistant Professor of Computer Science NC State University
Dr. Tiffany Barnes
Associate Professor of Computer Science at NC State University
Ph.D student of Computer Science at NC State University
Ph.D student of Computer Science at NC State University