🗺️ StuntingMap

Platform Intelijen Geospasial Stunting Indonesia

Peta risiko stunting interaktif mencakup 514 kabupaten/kota di seluruh Indonesia Data real-time dari SSGI, Riskesdas, BPS Susenas — ditenagai ML pipeline & narasi LLM

Overview

StuntingMap is a production-grade geospatial intelligence platform tracking child stunting risk across all 514 Indonesian districts (kabupaten/kota). Built as a senior AI engineer portfolio project demonstrating:

• Real data: SSGI 2022, Riskesdas 2018, BPS Susenas 2023 — no synthetic data
• ML models: XGBoost + SHAP for stunting prediction, PySAL for spatial autocorrelation
• LLM narratives: Pre-generated GPT-4o-mini district reports in Bahasa Indonesia
• Geospatial: MapLibre GL JS choropleth with 8 data layers + time slider (2013-2024)

National context: Indonesia's stunting prevalence was 19.8% in 2024, missing the 14% national target. The worst-affected province is NTT at 35.9%; the best is Bali at 10.9%.

---

📸 Screenshots

🗺️ Interactive Map — 8 Data Layers · Time Slider 2013–2024

📊 National Dashboard — Real-time KPIs · Provincial Rankings

---

🚨 National Context

Indonesia recorded 19.8% stunting prevalence in 2024 — missing the 14% national target by 5.8 percentage points.

Province	Prevalence	Status
NTT	35.9%	🔴 Worst
Papua	32.5%	🔴 Critical
Sulawesi Barat	31.5%	🔴 Critical
Maluku	29.2%	🟠 High
DKI Jakarta	10.2%	🟢 Best
Bali	12.8%	🟢 Below target
DIY Yogyakarta	13.5%	🟢 Below target

---

✨ Features

Feature	Description
🗺️ Choropleth Map	MapLibre GL JS, 8 data layers, per-district interactive tooltips
⏱️ Time Slider	Temporal animation 2013–2024, adjustable speed
📊 National Dashboard	KPI cards, progress gauge toward 14% target, provincial ranking
🔍 District Comparison	Side-by-side comparison across kabupaten/kota
🧠 ML Pipeline	XGBoost + SHAP, Moran's I, LISA spatial clustering
📝 LLM Narratives	GPT-4o-mini, pre-generated in Bahasa Indonesia
🎨 Colorblind-safe	YlOrRd (stunting/SRI) · RdYlGn (WASH/health)
📱 Responsive	Tablet-friendly, touch gestures supported
🔐 Auth	Full public access + researcher login for saved views

---

🏗️ Tech Stack

Layer	Technology
Frontend	Next.js 16 (App Router), TypeScript, Tailwind CSS
Map	MapLibre GL JS, GeoJSON (optimized <3MB)
Charts	Recharts, D3 (SVG gauges)
State	Zustand
Database	PostgreSQL + PostGIS (via Drizzle ORM)
Cache	Redis (ioredis)
ML Pipeline	Python: XGBoost, SHAP, PySAL, Pydantic v2
LLM	OpenAI GPT-4o-mini (pre-generated, cached)
Orchestration	Prefect
Package Manager	pnpm workspaces + Turborepo
Deploy	Vercel (web) + Railway/Docker (ML service)

---

📦 Data Sources

Dataset	Source	Coverage	License
SSGI 2022 (Survei Status Gizi Indonesia)	Kemenkes RI	34 provinsi, 514 kab/kota	Public
Riskesdas 2018	Badan Litbangkes	Nasional	Public
BPS Susenas 2023	Badan Pusat Statistik	Nasional	Public
HDX Indonesia Admin Boundaries	OCHA/HDX	Level 2 (kab/kota)	CC BY
PODES 2021	BPS	83,794 desa	Public

Transparency note: Where district-level data is unavailable for specific years, this platform uses ML models (XGBoost) to estimate values. All modeled values are clearly labeled as "Estimasi Model" versus measured "Data Terukur" throughout the interface.

---

🚀 Getting Started

Prerequisites

• Node.js ≥ 18, pnpm ≥ 9
• Docker Desktop (for PostgreSQL + Redis)
• Python 3.11+ (for ML pipeline)
• OpenAI API key (optional, for LLM narratives)

Installation

# Clone repository
git clone https://github.com/vaskoyudha/StuntingMap.git
cd StuntingMap

# Install dependencies
pnpm install

# Start infrastructure
docker compose up -d

# Copy environment variables
cp .env.example .env.local
# Edit .env.local with your credentials

# Setup database
pnpm db:push
pnpm db:seed

# Start development server
pnpm dev

The app will be available at http://localhost:3000

Environment Variables

DATABASE_URL=postgresql://stuntingmap:stuntingmap@localhost:5432/stuntingmap
REDIS_URL=redis://localhost:6379
OPENAI_API_KEY=sk-your-key-here          # Optional — template narratives used if missing
NEXT_PUBLIC_APP_URL=http://localhost:3000
ML_API_URL=http://localhost:8000
NEXTAUTH_SECRET=your-secret-here
PIPELINE_SECRET=your-pipeline-secret

---

📁 Repository Structure

StuntingMap/
├── apps/
│   ├── web/                        # Next.js 16 App Router
│   │   ├── app/                    # Pages + API routes
│   │   │   ├── map/               # Interactive map page
│   │   │   ├── dashboard/         # National analytics
│   │   │   ├── compare/           # District comparison
│   │   │   └── api/               # REST API (15 routes)
│   │   ├── components/
│   │   │   ├── map/               # MapLibre components (6)
│   │   │   ├── district/          # District panel (8)
│   │   │   ├── dashboard/         # Dashboard widgets (5)
│   │   │   └── shared/            # Reusable UI (4)
│   │   ├── lib/                   # DB, Redis, colors, utils
│   │   └── store/                 # Zustand state management
│   └── pipeline/                  # Python ML service
│       ├── analysis/              # Moran's I, LISA
│       ├── models/                # XGBoost, SRI, trajectory
│       ├── narratives/            # LLM generation
│       └── pipelines/             # Prefect orchestration
├── packages/
│   ├── db/                        # Drizzle schema (9 tables)
│   └── shared/                    # Shared TypeScript types
├── scripts/
│   ├── ingest/                    # Data ingestion (Python)
│   └── tiles/                     # GeoJSON generation
└── data/processed/                # Real Indonesian CSV data

---

🔌 API Reference

Endpoint	Method	Description
/api/geojson	GET	GeoJSON boundaries + stunting data
/api/layers	GET	Available data layers
/api/districts	GET	All districts with latest data
/api/districts/[pcode]	GET	Single district detail
/api/districts/[pcode]/report	GET	Full district report
/api/districts/[pcode]/peers	GET	Peer comparison districts
/api/national	GET	National-level statistics
/api/scatter	GET	Scatter plot correlation data
/api/v1/districts	GET	Public API v1

---

🧠 Methodology

Stunting Risk Index (SRI)

SRI is a composite index (0–1) combining:

• Prevalence score (30%): Current stunting prevalence normalized nationally
• Trend score (20%): Historical trajectory (improving vs worsening)
• Determinant score (30%): WASH access, poverty, education composite
• Intervention gap score (20%): Healthcare coverage gaps

Spatial Analysis

• Global Moran's I: Tests for spatial clustering of stunting across Indonesia
• LISA (Local Moran's I): Identifies HH/LL/HL/LH clusters at district level
• Weights: K-nearest neighbors (k=8), row-standardized, robust for archipelago geography

Scenario Projections

Three scenarios based on historical trend slope:

• BAU: Business as usual — continues historical trend
• Accelerated: 2.5× BAU decline with intensive intervention
• Minimal: 0.3× BAU decline with reduced resources

---

🧪 Testing

# TypeScript type checking
cd apps/web && pnpm tsc --noEmit

# Python ML pipeline tests (9/9 passing)
cd apps/pipeline && pytest tests/ -v

---

🚢 Deployment

Vercel (Frontend)

vercel deploy

Railway (ML Pipeline)

docker build -t stunting-pipeline apps/pipeline/
railway up

---

🤝 Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feat/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feat/amazing-feature)
5. Open a Pull Request

---

📜 License

MIT License — see LICENSE for details.

Data from SSGI, Riskesdas, and BPS Susenas is used under Indonesian government open data provisions. HDX boundaries under CC BY OCHA.

---

📖 Citation

If you use StuntingMap data or methodology in research, please cite:

@software{stuntingmap2024,
  title  = {StuntingMap: Platform Intelijen Geospasial Stunting Indonesia},
  author = {Nodyatama, Vasco Yudha},
  year   = {2024},
  url    = {https://github.com/vaskoyudha/StuntingMap},
  note   = {Data: SSGI 2022 (Kemenkes RI), Riskesdas 2018, BPS Susenas 2023}
}

---

Dibuat untuk mendukung penurunan stunting di Indonesia 🇮🇩

🗺️ Map · 📊 Dashboard · 🐛 Report Bug · 💡 Request Feature