🪐 Surface Temperature Predictor

Introduction

The project utilizes the Global Land Temperature By City dataset from Berkeley Earth, available on Kaggle, and employs the Meta AI Prophet Framework for temperature forecasting.

Key Features

• Dynamic Visualizations: Generate pie charts, geographic maps, and statistical summaries to explore temperature data interactively.
• Global and Regional Insights: Analyze temperature trends for cities worldwide and detailed U.S. state-level data.
• Forecasting Capabilities: Predict future temperature trends using machine learning models.
• Streamlit-Powered Interface: A simple and accessible web-based application for users of all technical backgrounds.
• Docker Support: Easily deploy the application using Docker for a consistent runtime environment.

Installation and Setup

Follow these steps to set up and run the World Temperature Viewer locally or in a containerized environment.

Prerequisites

• Python 3.8 or higher
• Docker (optional, for containerized deployment)
• Recommended: A virtual environment (e.g., venv or conda)

Local Setup

Clone the Repository

git clone https://github.com/your-repo/WorldTemperatureViewer.git
cd WorldTemperatureViewer

Install Dependencies

Install the required Python libraries:

pip install -r requirements.txt

Run the Application

Start the Streamlit application:

streamlit run Vizualisations.py

Access the Application

Once the application is running, open the provided local URL (e.g., http://localhost:8501) in your browser.

Docker Setup

Build the Docker Image

Navigate to the project directory and build the Docker image:

docker build -t world-temperature-viewer .

Run the Docker Container

Start the application in a Docker container:

docker run -p 8501:8501 world-temperature-viewer

Access the Application

Open your browser and navigate to http://localhost:8501 to use the application.

Folder Structure

The repository is organized as follows:

WorldTemperatureViewer/
├── data/                     # Contains sample datasets or links to external datasets
├── notebooks/                # Jupyter notebooks for exploratory data analysis
├── models/                   # Trained forecasting models and configurations
├── src/                      # Source code for data processing and visualization
├── Vizualisations.py         # Main Streamlit application file
├── requirements.txt          # Python dependencies
├── Dockerfile                # Docker configuration for containerized deployment
├── README.md                 # Project documentation

Dataset

The application uses the Global Land Temperature By City dataset, which can be downloaded from Kaggle: Berkeley Earth Dataset

Methodology

The dataset was divided into training (1970–2013) and testing (2010–2013) sets. The Prophet Framework was used for forecasting with the following parameters:

{
    'growth': 'linear',
    'seasonality_mode': 'additive',
    'seasonality_prior_scale': 10.0,
    'holidays_prior_scale': 10.0,
    'changepoint_prior_scale': 0.05,
    'mcmc_samples': 0,
    'interval_width': 0.8,
    'uncertainty_samples': 1000,
    'stan_backend': None
}

The model achieved an Average Mean Absolute Error (1.3) and Average Mean Squared Error (7.1), demonstrating robust performance.

Feedback and Contributions

We welcome your feedback and contributions to improve the project.

Thank you for supporting this initiative to promote climate awareness and sustainable practices!