🌊 RiskRadar — Flood Risk Prediction System

An End-to-End Machine Learning System for Flood Risk Prediction using Hydrological, Meteorological, and Geospatial Data.
Built using XGBoost, deployed with FastAPI, and visualized through an interactive Streamlit Dashboard.

---

---

🚀 Live Demo

👉 Try RiskRadar Live

---

📌 Table of Contents

• 📖 Overview
• ✨ Key Features
• 🧠 Model & Methodology
• 📊 Results
• 🏗️ Project Structure
• ⚙️ Installation
• 🚀 Running the Project
• 🔌 API Reference
• 🐳 Docker Support
• 🛠️ Tech Stack
• 📈 Future Scope
• 📜 License
• 👨‍💻 Author

---

📖 Overview

RiskRadar is a production-grade flood risk prediction system that analyzes hydrological, meteorological, and geospatial parameters to classify locations into flood risk categories.

The project leverages a high-performance XGBoost Machine Learning model trained on real-world flood datasets and provides predictions through:

• 🚀 FastAPI REST API
• 📊 Interactive Streamlit Dashboard
• 🐳 Dockerized Deployment Pipeline

Unlike a traditional ML notebook project, RiskRadar focuses on a complete deployment-ready architecture suitable for real-world usage.

---

✨ Key Features

✅ Trained on 45,932 global flood locations
✅ Achieved 96.9% Accuracy
✅ F1 Score: 0.984
✅ Uses SMOTE for class imbalance handling
✅ Real-time prediction dashboard with Plotly visualizations
✅ FastAPI backend with input validation
✅ Interactive flood-risk mapping
✅ Docker-ready deployment
✅ Production-oriented ML workflow

---

🧠 Model & Methodology

📂 Dataset

• Source: IIT Delhi Static Flood Database
• Dataset File: kuntla_flooddatabase.csv
• Raw Features: 93
• Final Engineered Features: 13
• Target: Flood Risk Classification

---

🔄 Machine Learning Pipeline

Raw Dataset (93 Parameters)
        ↓
Missing Value Handling
        ↓
Feature Engineering
        ↓
Categorical Encoding
        ↓
Train/Test Split
        ↓
SMOTE Balancing
        ↓
Hyperparameter Tuning
        ↓
XGBoost Training
        ↓
Evaluation & Model Export
        ↓
FastAPI + Streamlit Deployment

---

⚙️ Data Preprocessing

Missing Value Handling

• Dropped columns with more than 40% missing values
• Median imputation for numerical features

Feature Engineering

Created custom engineered features such as:

• precipitation_risk_index
• drainage_risk_index

Encoding

Applied Label Encoding for:

• Climate Type
• Landcover Type
• Soil Type
• Lithology Type

---

🤖 Model Training

Algorithm Used

• XGBoost Classifier

Hyperparameter Optimization

• RandomizedSearchCV
• 50 parameter combinations
• 3-fold cross-validation
• GPU accelerated training (tree_method=hist)

Handling Imbalanced Data

• Applied SMOTE on training dataset

---

📊 Results

Metric	Score
Accuracy	96.9%
F1 Score	0.984
Recall	0.791

---

🌍 High Risk Locations

• High-risk locations identified: 2,538
• Total locations analyzed: 45,932

---

🔥 Most Important Risk Factors

• Annual Precipitation
• Seasonal Rainfall Patterns
• Drainage Density
• Basin Relief
• Soil Characteristics
• Land Cover Features
• Temperature Seasonality

---

🏗️ Project Structure

riskradar/
│
├── flood_fast.py
├── api.py
├── app.py
├── Dockerfile
├── requirements.txt
├── flood_model.pkl
├── encoders.pkl
├── feature_columns.pkl
├── high_risk_locations.csv
└── README.md

---

⚙️ Installation

1️⃣ Clone Repository

git clone https://github.com/nitinc264/Flood-prediction-Model.git
cd Flood-prediction-Model

---

2️⃣ Create Virtual Environment

Windows

python -m venv venv
venv\Scripts\activate

Linux / Mac

python3 -m venv venv
source venv/bin/activate

---

3️⃣ Install Dependencies

pip install -r requirements.txt

---

🚀 Running the Project

▶️ Start FastAPI Backend

uvicorn api:app --reload --port 8000

API Documentation

http://localhost:8000/docs

---

▶️ Start Streamlit Dashboard

Open a new terminal and run:

streamlit run app.py

Dashboard URL

http://localhost:8501

---

🔌 API Reference

Method	Endpoint	Description
GET	/	API Information
GET	/health	Health Check
GET	/features	Feature List
POST	/predict	Flood Risk Prediction

---

📥 Sample Prediction Request

curl -X POST http://localhost:8000/predict \
-H "Content-Type: application/json" \
-d '{
  "annual_precipitation": 1500,
  "precipitation_of_wettest_month": 400,
  "precipitation_seasonality": 60,
  "drainage_density": 3.5,
  "drainage_texture": 12.0,
  "basin_relief": 800,
  "annual_mean_temperature": 22,
  "temperature_seasonality": 600,
  "curve_number_amcii": 75,
  "ruggedness_number": 0.8,
  "infiltration_number": 6,
  "climate_type": 2,
  "landcover_type": 1,
  "soil_type": 3
}'

---

📤 Sample Response

{
  "flood_risk_score": 0.8341,
  "risk_percentage": "83.4%",
  "risk_level": "HIGH",
  "color": "red",
  "message": "Immediate attention required. High probability of flood event.",
  "features_used": 13
}

---

🐳 Docker Support

Build Docker Image

docker build -t riskradar .

---

Run Docker Container

docker run -p 8000:8000 riskradar

---

🛠️ Tech Stack

Layer	Technology
Machine Learning	XGBoost, Scikit-learn
Data Processing	Pandas, NumPy
Imbalanced Learning	SMOTE
Backend API	FastAPI, Uvicorn
Frontend Dashboard	Streamlit
Visualization	Plotly, Folium, Matplotlib
Deployment	Docker
Training Environment	Google Colab

---

📈 Future Scope

• Hugging Face deployment ✓ — Live at nitin454545-riskradar.hf.space
• Real-time weather API integration
• Global-scale flood dataset expansion
• SMS & Email flood alerts
• Mobile-responsive frontend
• Live geospatial monitoring
• Cloud deployment pipeline

---

📜 License

This project is licensed under the MIT License.

You are free to:

• Use
• Modify
• Distribute
• Improve

with proper attribution.

---

👨‍💻 Author

Nitin Chauhan

🎓 B.Tech — AI & Data Science
🏫 Dr. D.Y. Patil School of Science and Technology

---

⭐ If you found this project useful, don't forget to star the repository!