3. Data
• Demographics (age, gender, country,
education level)
• Timing (course launch, student registration,
student drop out, # distinct days worked)
• Certification in one or more courses
4. Predictive Model
Trained data with different classifiers and
cross-validated. Best model is Quadratic
Discriminant Analysis, with 75% prediction
accuracy.
5. Most relevant features
Demographics and previous
experience not as important as:
• How late is the student joining the course?
• How many days per week is she/he
willing to work?
• What fraction of the course will be
completed by the student?