Seminar on Educational Data Mining and Learning Analytics
3
Algoritmit ja koneoppiminen
Syventävät opinnot
Educational Data Mining (EDM) and Learning Analytics (LA) are research fields that focus on developing methodologies and tools to explore and analyze data gathered from educational settings. The roots of EDM are in constructing, studying and improving intelligent tutoring systems that guide and teach students, while LA has roots in the analysis of data originating from different types of learning management systems with the goal of improving offered education. Methods and practices in both fields are very similar, and various forms of machine learning and data mining practices are in the very core of both fields.
| Vuosi | Lukukausi | Päivämäärä | Periodi | Kieli | Vastuuhenkilö |
|---|---|---|---|---|---|
| 2016 | kevät | 22.01-06.05. | 3-4 | Englanti | Arto Hellas |
Luennot
| Aika | Huone | Luennoija | Päivämäärä |
|---|---|---|---|
| Pe 10-12 | C221 | Arto Vihavainen | 22.01.2016-04.03.2016 |
| Pe 10-12 | C221 | Arto Vihavainen | 18.03.2016-06.05.2016 |
Information for international students
This seminar will be held in English.
Yleistä
We focus on both research and practice papers in recent EDM and LA conferences and relevant journals.
Learning objectives
- Improve scientific and technical writing skills
- Improve scientific and technical presentation skills
- Learn about educational data mining and learning analytics
Course schedule:
- 29.1. (Topic: Hint Generation) Discussing articles (1) Experimental Evaluation of Automatic Hint Generation for a Logic Tutor and (2) An Exploration of Data-Driven Hint Generation for an Open-Ended Programming Problem, which everyone have read by the meeting. Virpi acts as the supporting opponent for (1), while Ada acts as the supporting opponent for (2). Possibility to already select topics for abstract and seminar report.
- 12.2. (Topic: Adaptive Learning Systems) Discussing articles (The Computer as Coach: An Athletic Paradigm for Intellectual Education, Mikko) and (Exploring the Assistance Dilemma in Experiments with Cognitive Tutors, Angel). Check out also the page at http://www.educationdive.com/news/adaptive-learning-the-best-approaches-weve-seen-so-far/187875/. Topics selected.
- 26.2. (Topic: Predicting Students' Performance) Discussing articles (Modeling How Students Learn to Program, Juho) and (Cluster-Based Prediction of Mathematical Learning Patterns, Janne). Abstracts written and submitted.
- 4.3. Writing the seminar report starts at the latest. No meeting.
- 11.3. Abstracts reviewed. Voluntary meeting, discussing seminar report structure, formatting, content, ...
- 18.3. (Topic: Modeling Students' Knowledge) Discussing articles (Using a Learning Agent with a Student Model, Adnan), (Tutor Modeling Versus Student Modeling, Arto) and (Identifiability: A Fundamental Problem of Student Modeling, Ada). Short pitches.
- 8.4. (Topic: Replicability and generalization, Educational games) Discussing articles (The use of computer games as an educational tool: Identification of game types and appropriate game elements, Adnan), (Practical, Appropriate, Empirically-Validated Guidelines for Designing Educational Games, Angel) and (Educational Data Mining and Learning Analytics in Programming: Literature Review and Case Studies, Sections 3 and 5, Mikko), drafts of the seminar paper submitted.
- Note! Extension: Drafts of the seminar paper submitted on 22.4.
- Note! 22.4. Session cancelled -- no discussion / articles.
- 2.5. Seminar reports submitted. Seminar reports should be 10-15 pages using the template at https://github.com/UniversityHelsinkiTKTL/tktltiki2.
- 6.5. Seminar presentations at D234. There will be a computer with an internet connection, where you can download PDFs for presentation. Each presentation should be around 10-15 minutes, after which there will be time for questions. Overall, the presentation should be on a level that can be followed by anyone who has taken the EDM seminar. The presentations will start at 9:30.
- 15.5. Seminar reports reviewed.
Kurssin suorittaminen
Participants must have completed the course Scientific writing or have equivalent skills. Students complete this seminar by actively participating in its work: the work methods include studying scientific sources, writing reports and giving presentations, reading the reports of other participants and evaluating them, and actively following presentations.
The grading will be based on each student's own written work (1/2), oral presentations (1/4), and commentary on the reports of others as well as activeness in general (1/4). To pass the seminar, each of these components must be passed. Active attendance of seminar meetings is mandatory.
During the course, you will (1) write an abstract describing your topic, (2) give a short "elevator pitch"-type presentation based on the abstract, (3) participate in discussions on relevant papers in the course, (4) write a seminar report, and (5) give a "conference-style" presentation that is based on your report.
The seminar meets for the first time on Jan. 22nd at 10 AM. During the meeting, we will discuss the primer articles, the seminar schedule (bi-weekly meetings, seminar day for conference presentations), and outline possible seminar report topics.
There is an IRC channel for course participants on IRCnet -- #edmla
Forums
- Educational Data Mining
- Learning at Scale
- Learning Analytics and Knowledge
- Intelligent Tutoring Systems
- User Modeling, Adaptation and Personalization
There are also a few journals and special issues which may be of interest -- perhaps JEDM can be a starting point.
Possible topics and keywords to search for
- Knowledge tracing / Bayesian knowledge tracing / Modeling Students' Knowledge, Mikko
- Learning factors analysis / Performance factor analysis, Juho
- Mastery learning and Wheel-spinning
- Detecting Student Misuse of ITS systems / Gaming the system / "Externalizing Confidence to the Feedback System", Virpi
- Automatic hint generation (in programming), Adnan
- Relationships between tasks and required knowledge (e.g. Q-Matrix)
- Predicting course outcomes / Predicting Student Dropouts (see e.g. 1), Angel
- Characterizing students
- Analyzing open-ended text responses (e.g. 1 -- accessible through Uni network)
- Language Processing Specific Topic (Relevant and Irrelevant Sentences in Math Problems), Ada
- Biometric feedback in pair programming, Arto
- Deciding when and which questions to give out to the students (e.g. using multi-armed bandits), Janne
- your own topic