Week 5: Educational Data Mining
Bodong Chen
Feb 23, 2015
SIGs and WGs
Ready to roll?? Yeah!
- Plan in advance
- Meet with Bodong one week in advance to finalize plans
(Updates from the CSCL + LA workshop)
Readings
- Scheuer, O. and McLaren, B. M. (2012). Educational data mining (or link 2). In Encyclopedia of the Sciences of Learning, pages 1075–1079. Springer.
- Baker, R.S.J.d., Yacef, K. (2009) The State of Educational Data Mining in 2009: A Review and Future Visions. Journal of Educational Data Mining, 1 (1), 3-17.
Issues in discussion
- Statistics vs. Data Mining
- EDM vs. LA
EDM (paper 1 & 2)
Educational Data Mining is concerned with developing, researching, and applying computerized methods to detect patterns in large collections of educational data – patterns that would otherwise be hard or impossible to analyze due to the enormous volume of data they exist within.
Educational Data Mining is an emerging discipline, concerned with developing methods for exploring the unique types of data that come from educational settings, and using those methods to better understand students, and the settings which they learn in.
(vs. LA?)
Typical steps in an EDM project (paper 1)
- data acquisition
- data preprocessing
- data mining
- validation of results
(vs. LA?)
Methods in EDM (paper 1)
- Prediction
- Clustering
- Relationship mining
- Distillation of data for human judgment
- Discovery with models
Application Areas
Bienkowski, 2012 (Week 2)
- User knowledge modeling
- User behavior modeling
- User experience modeling
- User profiling
- Domain modeling
- Trend analysis
- Adaptation and personalization
Application Areas (paper 1)
- Scientific inquiry and system evaluation
- Determining student model parameters
- Informing domain models
- Creating diagnostic models
- Creating reports and alerts for instructors, students and other stakeholders
- Recommending resources and activities
Application Areas (paper 2)
- Statistics and visualization
- Web mining
- Clustering, classification, and outlier detection
- Association rule mining and sequential pattern mining o Text mining
Terms
- machine learning
- text mining
- psychometrics
- web log analysis
- student model
- supervised vs. unsupervised learning
- over-fitting vs. under-fitting
- cross-validation
- relationship mining
- feature engineering
- A/B testing
- …
Feature engineering
Data Science?
Statistics vs Data Mining
| Statistics | Data Mining |
|---|---|
| Math is important | Yep, math is important |
| Model-based (theoretical driven) | Ad hoc (for a particular purpose) |
| A method/model that should work prior to its use | Experimental, ongoing refinement |
| Rigor is key | 'Adventurous' to some degree |
| Model is 'king' | Criteria of picking features is key |
| Smaller data – sample->population inference | Large data – could cover the population |
| Cleaner data | Messy data – data wrangling and cleansing |
| Numeric | Various forms of data |
| Confirmatory (mostly) | Exploratory (esp. with big dataset) |
| Generalization ability is important | Model 'fit' is important |
| Algorithms are less central | Algorithms are central |
| Data–analyst interaction | Data–(super-)computer–analyst interaction |
| Less likely to be real-time | Real-time in many cases |
Week 7: Cases and Examples
Find cases, share & discuss them in KF
- highlight 'awesomeness' :)
- connect with readings
Guest speaker: Vitomir Kovanovic, University of Edinburgh
- When: 4:40PM (as usual)
- Where: Google Hangouts