Flood Risk Prediction System 🌊

Overview

This project is a comprehensive AI and Machine Learning-based application for predicting flood risks in specific regions (Mostar, Jablanica, and Fojnica). By utilizing historical and real-time weather data, the system evaluates potential flood risks and offers valuable insights.

---

Features

1. Machine Learning Model

• Algorithm: Random Forest Regressor from scikit-learn.
• Input Features:
  • Temperature: temp
  • Humidity: humidity
  • Precipitation: precip
  • Precipitation Probability: precipprob
  • Wind Speed: windspeed
• Target: Predict precipitation levels.
• Automatic Retraining: The model dynamically updates with new data to enhance prediction accuracy.

2. Backend API

• Built using Flask with endpoints for:
  • Flood risk prediction.
  • Retraining the model with updated data.
• Uses Flask-CORS to enable cross-origin requests.
• Fetches real-time weather data via the Visual Crossing Weather API.

3. Interactive GUI

• Developed using Tkinter.
• Features:
  • Select a city (Mostar, Jablanica, or Fojnica).
  • Predict flood risks for up to 10 days ahead.
  • Displays results in a user-friendly, tabular format.

4. Data Processing

• Historical weather data merged and processed using pandas.
• Combines data from multiple CSV files for model training.
• Handles missing or invalid data through preprocessing.

---

Technologies Used

Programming Languages and Frameworks

• Python: Core programming language.
• Flask: For building the backend API.
• Tkinter: For creating the desktop GUI.

Machine Learning

• scikit-learn:
  • Used to implement the Random Forest Regressor.
  • Saves and loads the trained model using joblib.

Data Processing

• pandas: For data manipulation and merging CSV files.
• NumPy: Efficient handling of numerical data.

API Integration

• Visual Crossing Weather API: Fetches real-time weather data based on city and date.

Other Tools

• joblib: For saving and loading trained machine learning models.
• Flask-CORS: To manage cross-origin requests.

---

Installation

Prerequisites

1. Python 3.8+
2. Required Python packages (install via pip): pip install pandas scikit-learn flask flask-cors joblib requests
3. Access to the Visual Crossing Weather API (you will need an API key).

Steps to Run

1. Clone the repository: git clone https://github.com/Selma-Bajramovic/ai-agent-flood-prediction.git cd flood_prediction
2. Train the model: python model.py
3. Start the Flask backend: python app.py
4. Launch the Tkinter GUI: python frontend.py

---

File Structure

├── backend
│   ├── data
│   │   ├── fetch_data.py
│   │   ├── fojnica_weather.csv
│   │   ├── jablanica_weather.csv
│   │   ├── merged_data.py
│   │   ├── merged_flood_data.csv
│   │   └── mostar_weather.csv
│   ├── model
│   │   └── flood_prediction_model.pkl
│   ├── app.py
│   └── model.py
├── frontend
│   └── frontend.py
└── README.md

---

How It Works

1. Data Collection: -Historical weather data is merged from multiple CSV files. -Real-time weather data is fetched using the Visual Crossing Weather API.
2. Training the Model: -The model is trained using the RandomForestRegressor on historical data. -Input features include weather parameters like temperature and precipitation.
3. Flood Risk Prediction: -The Flask API uses the trained model to predict precipitation levels. -Risk levels (Low, Moderate, High) are assigned based on thresholds.
4. Interactive GUI: -Users input the city and forecast days. -Predictions are displayed in a detailed, interactive table.

---

Acknowledgments

• Visual Crossing Weather API: For providing real-time weather data.
• scikit-learn: For the robust machine learning library.
• Inspiration and feedback from educational and professional projects.

🌟 Thank you for exploring this project!