Educational Data Mining WS 07/08

• the two introductory courses will take place Monday 29th of October in Seminarraum 016 and 5th of November in a location to be announced
• the presentations will take place on 28th and 29th of February 2008

Abstract

English: The seminar introduces Educational Data Mining (EDM). Educational data mining is the process of converting raw data from educational systems to useful information that can be used to better understand and assess students and the contexts which they learn in, inform the design and improve the performance of educational systems, answer research questions and inform teachers and researchers.
EDM encompasses a wide range of research techniques that includes more traditional options such as database queries and simple automatic logging as well as more recent developments in machine learning, data mining and language technology.
EDM is of particular interest now due to the scaling up of the number of students using interactive learning environments such as intelligent computer tutors. When the field of computer-based education was new, the main challenge of student modeling was to build a basic model of the students' competencies. Few students used such systems, and controlled studies were of brief duration and had relatively few users. In recent years, studies involving computer tutors have scaled up in scope both longitudinally and in the number of users. This increase in scale has created a problem: what to do with the data? For the first time we have the ability to answer educational questions about how individual students will react to instruction, or whether a particular student is not learning material in an expected manner. The missing ingredient is the computational toolkit to organize, visualize, and learn from the data. For instance, researchers have used educational data mining to:

• Detect affect and disengagement
• Detect attempts to circumvent learning called "gaming the system"
• Guide student learning efforts
• Develop or refine student models
• Measure the effect of individual interventions
• Improved teaching support
• Predict student performance and behavior

Structure of seminar

• Introduction to Machine Learning Techniques
• ActiveMath description and demo
• Technology, data formats and granularity and content of data
• EDM for metadata estimation
• EDM for student modelling (several presentations)
• EDM to guide student learning efforts
• EDM for building ITS
• Learning from collaboration data
• Online mining for teacher support
• Text mining for classifying student inputs

Grading criteria

Presence is expected in all sessions.
Deadlines for first version of slides 2 weeks before seminar.
At least 2 meetings with supervisor are expected.

• 70% - Quality of presentation of article and related material
• 10% - Presence, deadlines for first version of slides, at least two meetings with supervisor
• 10% - Quality and quantity of active participation in live and online discussion
• 10% - Quality and quantity of individual ePortfolio

• You may make handouts, but this is rather unusual.
• Speak to the audience, never turn your back to the audience to look at your slides.
• Don't present your slides while reading your text from a pice of paper aloud.
• Keep time constraints.
• 15 slides on average for a 20 min presentation are okay.
• Use keywords. Omit sentences as much as possible. Your audience will start to read your slides and miss important information you are saying.
• Keep the focus on the key idea.
• However, you are allowed to add as many slides as you like.
  • You can put these eitehr after your "Thank you." as a reserve just in case someone asks questions. Then you can always refer to these and it shows the audience that you have that competency and that you moved that slide to the back due to time constraints.
  • You can embed the slides in your talk but you can skip them if you think you get in a time problem. You say:"I skip this and go on with more relevant slides."
• We are computer scientists. Explain algorithms!
• Show how different components integrate/communicate with each other. Show the big picture.
• If you can explain something by an illustration, don't use text. Pictures are much more reliable.
• Show your transfer knowledge. Explain, how you would apply the presented approach to ActiveMath. What are the prerequisites?

Some questions (examples) to be discussed:

• How can data mining be used to improve the ActiveMath system, interventions and content?
• Which data need to be logged for those improvements? Which data to store and to mine? which level of granularity?
• How can we integrate data mining and existing educational theories? What techniques are especially useful for data mining?

Prerequisites

• Basic Artificial Intelligence Knowledge, Machine Learning
• Knowledge of Educational Technology would be useful but is not strictly required

Possible follow-up activities

You can make your bachelor and master on ActiveMath-related topics including educational data mining. Have a look at our web page or get in touch with us.