edm-learning-analytics

You can forecast student dropout risk by fitting Cox survival models on normalized LMS event logs to predict time-to-dropout and quantify hazard ratios using survival curves. The process requires extracting early engagement and assessment data from platforms like Moodle or Canvas to identify at-risk students.

Bayesian Knowledge Tracing is an educational data mining method that estimates the per-student probability of knowing a specific skill over time. It calculates posterior mastery estimates from attempt-level correctness data in quiz and tutoring logs to track learning progress.

You can discover frequent activity sequences in learning behavior data by applying sequence mining techniques to normalized student-time-event datasets. This process identifies behavioral patterns from LMS interaction logs that correlate with academic success or struggle.

Yes, you can estimate student abilities and item difficulties by fitting a 1PL Rasch IRT model to assessment response data. This educational data mining approach processes quiz and tutoring results to evaluate both learner skill levels and question complexities.

Yes, you must construct normalized student-time-event datasets from your Moodle or Canvas CSV exports before analysis. Proper data formatting structures the raw LMS interaction logs and assessment results required for fitting BKT, Cox, and IRT models.

Educational data mining workflows require numpy, pandas, scipy, scikit-learn, lifelines, and matplotlib. These libraries support fitting Bayesian Knowledge Tracing and Cox survival models, generating visualizations, and producing posterior mastery and hazard summaries.