July 14, 2024
09:00 AM - 9:30 AM: Features of components as predictors (Phil and Luke)
- Basis of LKT models: features for data components
- Computing features for levels in identifier columns per student
- Requirements for running models and preprocessing
- Time-based features and time-formatted data
- Assistance with data formatting and special cases
- Review of LKT model varieties (AFM, PFA)
- Examination of LKT function inputs, outputs, and feature examples
- Creation and fitting of AFM, PFA models with recency feature
10:30 AM - 11:00 AM: Break
11:00 AM - 11:30 AM: Searching for optimal feature & Crossvalidation (Luke)
- Introduction of stepwise and LASSO feature search methods
- Comparison of LKT models to deep learning approaches
- Crossvalidation of complex models using LKT functions
11:30 AM - 12:00 PM: Creating new features (Phil)
- Hands-on tutorial on adding features to LKT codebase
- Feature engineering using LKT
- Open-source contributions to the R package on CRAN
- Example of using fitted LKT model for practice scheduling
- Discussion of model features' influence on interpretation and ITS use
- Demonstration of using knowledge tracing model for student grouping
- Transforming model outputs into category labels
- Future directions in knowledge tracing