sameerbanchhor-work/durg-EDU-ai-models

🎓 DURG University Student Prediction Models (DURG-EduAI)

Models Included

File	Purpose	Accuracy
xgb_sgpa_model.pkl	SGPA Regression	R²=0.9969, MAE=0.079
xgb_result_classifier.pkl	PASS/FAIL/ATKT	99%+ F1
xgb_dropout_risk.pkl	Dropout Risk (Low/Med/High)	100% F1
subject_benchmarks.pkl	Per-subject cohort stats	161 subjects
artifacts.pkl	Feature fill values & encodings	—

Training Data

• Sources: MSC, MA, Other PG, UG (Hemchand Yadav University, Durg)
• PG records (SGPA): 129,833
• Classification records: 222,009
• Dropout risk records: 245,913

Usage

1. Upload sgpa_model/ folder to your Colab/environment
2. Load models with joblib.load()
3. Call analyze_student(record, source='msc'|'ma'|'other_pg'|'ug')
4. Call print_report(report) to display results

Predictions

1. SGPA (PG only) — regression, ±0.08 average error
2. Result Status — PASS / ATKT / FAIL with probabilities
3. Dropout Risk — Low / Medium / High with probabilities
4. Subject-wise analysis — score vs cohort average
5. Early Warning flags — automatic alerts for at-risk students

Source Encoding

msc=0, ma=1, other_pg=2, ug=3