🏦 Loan Portfolio Health Monitor

BA Portfolio Project – Executive dashboard with early warning system for NPA prevention
Stack: Python · Streamlit · SQLite · Plotly

---

🚀 Quick Start

# 1. Clone / place files in a folder
cd loan_monitor

# 2. Install dependencies
pip install -r requirements.txt

# 3. Generate synthetic data (5,000 loans)
python generate_data.py

# 4. Run the dashboard
streamlit run app.py

The app auto-generates data on first launch if data/loan_data.db is missing.

---

📁 Project Structure

loan_monitor/
├── app.py              ← Main Streamlit dashboard
├── generate_data.py    ← Synthetic data generator
├── requirements.txt    ← Python dependencies
├── data/
│   ├── loans.csv
│   ├── monthly_payments.csv
│   └── loan_data.db    ← SQLite database (auto-created)
└── README.md

---

📊 Dashboard Sections

Section	What It Shows
1. Overview KPIs	AUM, Gross NPA%, Net NPA%, Portfolio Yield, Collection Efficiency
2. Delinquency Funnel	Flow from Current → 30 DPD → 60 DPD → 90+ DPD with roll rates
3. Segmentation	NPA% by loan type, ticket size, and vintage
4. Early Warning Alerts	At-risk accounts before they become NPAs, with risk scores
5. Predictive Flag Engine	Rule-based flagging: delayed payments, partial EMIs, segment risk

---

🧮 Key Metrics Explained

Metric	Formula	RBI Benchmark
Gross NPA%	NPA Outstanding / Total AUM	< 5%
Net NPA%	(NPA − Provisions) / Net Assets	< 2%
Portfolio Yield	Weighted avg interest rate	Target > 12%
Collection Efficiency	Collected / Expected × 100	> 95%

---

⚡ Early Warning Rules

Three rule-based flags with weighted risk scores:

1. Payment Delayed 5+ Days (last 2 months) → 40 pts
2. Partial EMI Payment (not full EMI) → 35 pts
3. Segment with High Historical Default → 25 pts

Accounts scoring ≥ 70 = Critical (red)
Accounts scoring 40–69 = Watchlist (amber)

---

Synthetic data — for demonstration purposes only