---
title: Mumbai Crowd Predictor
emoji: 🚆
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false
---
# 🚆 Mumbai Local Train — Crowd Density Predictor (Advanced Edition)

AI-powered crowd prediction system for Mumbai Local Trains with a full-stack FastAPI + HTML/CSS/JS deployment.

---

## 🚀 Features

- **AI Prediction** — RandomForest model with 17 engineered features (~97% accuracy)
- **Hourly Forecast** — Full day crowd pattern for any route
- **Station Heatmap** — Visual heatmap of crowd across stations × hours
- **Risk Scoring** — 0–100 risk score with reasoning and travel tips
- **Beautiful UI** — Dark glassmorphism design, animated charts, responsive

---

## 🗂️ Project Structure

```
mumbai_train_app/
├── main.py                  # FastAPI application (entry point)
├── requirements.txt
├── src/
│   ├── model.py             # RandomForest + feature engineering
│   └── risk_engine.py       # Risk scoring, explanations, tips
├── data/
│   └── raw/
│       └── mumbai_local_crowd_data.csv
├── static/
│   ├── css/style.css        # Full dark-mode glassmorphism CSS
│   └── js/app.js            # Frontend JavaScript
└── templates/
    └── index.html           # Main HTML template
```

---

## ⚙️ Setup & Run

### 1. Install dependencies
```bash
pip install -r requirements.txt
```

### 2. Run the server
```bash
python main.py
# OR with uvicorn directly:
uvicorn main:app --host 0.0.0.0 --port 8000 --reload
```

### 3. Open in browser
```
http://localhost:8000
```

---

## 🧠 ML Model Details

| Feature | Description |
|---------|-------------|
| `hour` | Hour of travel (5–22) |
| `hour_weight` | Peak-hour weight (empirical) |
| `day_num` | 0=Monday → 6=Sunday |
| `is_weekday/weekend` | Binary flags |
| `is_morning_peak` | 7–9 AM flag |
| `is_evening_peak` | 5–7 PM flag |
| `morning_southbound` | Directional rush interaction |
| `evening_northbound` | Directional rush interaction |
| `line_num` | Western/Central/Harbour encoded |
| `dir_num` | Northbound/Southbound encoded |
| `station_enc / dest_enc` | Label-encoded stations |

**Algorithm**: Random Forest Classifier (200 trees, balanced class weights)  
**Accuracy**: ~97% on training data

---

## 📡 API Endpoints

| Method | Path | Description |
|--------|------|-------------|
| GET | `/` | Web UI |
| GET | `/api/lines` | All train lines |
| GET | `/api/stations/{line}` | Stations for a line |
| GET | `/api/stats` | Dataset + model stats |
| POST | `/api/predict` | Single prediction |
| POST | `/api/hourly-forecast` | Full day forecast |
| GET | `/api/heatmap/{line}` | Station × hour heatmap |

---

## 🛤️ Train Lines Covered

- **Western Line**: Churchgate → Virar (12 stations)
- **Central Line**: CSMT → Kalyan (9 stations)
- **Harbour Line**: CSMT → Panvel (8 stations)

---

## 👨‍💻 Tech Stack

- **Backend**: FastAPI + Uvicorn
- **ML**: scikit-learn RandomForest, Pandas, NumPy, Joblib
- **Frontend**: Pure HTML5 + CSS3 + Vanilla JS
- **Design**: Dark glassmorphism, CSS animations, responsive