AI-powered food waste prediction and environmental impact intelligence
What it does: Predicts food waste per student based on 20 operational factors
How it was trained: 50,000 simulated realistic school scenarios with noise (like real data)
Accuracy: Predicts waste within ±0.04 lbs/student on test data (80% R² score)
What it is NOT: A perfect forecast. Real schools have unexpected events, special meals, and variability.
Complex menus (many items, prep steps) lead to more waste. Simple 5-item menus reduce waste.
Lower attendance = more food prepared than eaten. Mondays & Fridays have lower attendance.
Larger portions = more waste. 9-10 oz portions have 15-20% less waste than 12-14 oz.
Understaffed kitchens = more waste. Adequate staffing improves quality control.
Poor equipment, bad weather, exams, holidays also impact waste.
Use these predictions to:
Do NOT use for:
❌ Does not account for:
⚠️ May not be accurate for:
📊 Data Quality: Model trained on realistic but simulated data. Real waste varies due to unmeasured factors (student moods, viral videos, social trends, etc.).
All 20 input factors are visible. You can see exactly what drives predictions.
Model uses only school-level operational data. No student names, IDs, or dietary restrictions.
Recommendations prioritize food rescue (feeding those in need) over just waste reduction.
After each prediction, see which factors had biggest impact on that specific result.
Get specific, feasible interventions—not vague recommendations.