All work

ML platform · 2024

Dropout Prediction & Counseling

Early-warning system for student dropout risk

Prediction accuracy
87%
Students monitored
200+
Mentor alerting
Automated

The problem

Institutions only discovered at-risk students after attendance and grades had already collapsed — far too late for counselling to help.

What we built

An ensemble model scores risk continuously from attendance, assessment and engagement signals, and Celery workers push alerts to mentors the moment a student crosses a threshold.

Key capabilities

  • Ensemble scoring

    Logistic Regression + Decision Tree ensemble at 87% accuracy.

  • Continuous monitoring

    Risk recalculated as new attendance and marks land.

  • Automated alerts

    Celery workers notify mentors and counsellors on threshold breach.

  • Counselling tracker

    Intervention notes tied to each student's risk timeline.

Next projects