Supplier Risk Intelligence Platform
Live App:
https://supplier-risk-intelligence.streamlit.app
An end-to-end machine learning system that predicts late deliveries, order cancellations, and profit margin risk — aggregated into a composite supplier risk score for operational decision-making.
Business Problem
Supply chain organizations struggle with:
- Late deliveries impacting SLA compliance
- Order cancellations reducing revenue reliability
- Low-margin orders increasing financial risk
- Lack of a unified supplier risk visibility framework
This platform builds predictive models and aggregates them into a composite risk score to enable:
- Proactive supplier monitoring
- Risk-based procurement prioritization
- Margin protection strategies
- Executive-level supplier risk dashboards
System Architecture
Raw Orders (180K rows)
↓
Feature Engineering (behavioral + historical)
↓
Model 1 → Late Delivery Classifier (XGBoost)
Model 2 → Cancellation Risk Classifier (XGBoost)
Model 3 → Profit Risk Model
↓
Weighted Composite Risk Score
↓
Supplier-Level Aggregation
↓
Streamlit Risk Intelligence Dashboard
Models & Performance
1️⃣ Late Delivery Classifier
- Algorithm: XGBoost
- ROC-AUC: ~0.73
- Optimized for balanced recall & precision
- Feature importance validated with SHAP
2️⃣ Cancellation / SLA Breach Model
- Handles severe class imbalance (~5.6% positive class)
- ROC-AUC: ~0.82
- Threshold tuning for recall-focused risk detection
3️⃣ Profit Risk Model
- Profit margin outlier clipping
- Engineered profit risk score
- Aggregated into supplier-level financial exposure signal
Composite Risk Score
Final supplier risk score combines:
- 50% Late Delivery Risk
- 30% Cancellation Risk
- 20% Profit Margin Risk
Suppliers are categorized into:
🔴 Critical
🟠 High
🟡 Moderate
🟢 Low
Dashboard Capabilities
- Executive risk tier summary
- Supplier rankings
- Department risk heatmap
- Risk contribution breakdown
- Drill-down analytics by category
- Downloadable risk report (CSV)
Tech Stack
- Python
- Pandas / NumPy
- Scikit-learn
- XGBoost
- SHAP
- Matplotlib / Seaborn
- Streamlit
- Git / GitHub
Dataset
DataCo Smart Supply Chain Dataset
180,519 orders (2015–2017)
Note: Raw dataset excluded from repository for size compliance.
Deployment
The application is deployed via Streamlit Cloud and auto-builds from the main branch.
Why This Project Matters
This project demonstrates:
- End-to-end ML system design
- Class imbalance handling
- Feature engineering strategy
- Multi-model aggregation
- Business-aligned scoring logic
- Production deployment workflow
- Dashboard UX for executive stakeholders
Author
Tejas Jaggi
Machine Learning & Supply Chain Analytics