rshewatkar/ship_co2-emission-predictor

🚢 Ship CO₂ Emission Prediction

📌 Project Overview

This project predicts CO₂ emissions of ships based on key operational and fuel-related parameters such as engine efficiency, fuel consumption, distance, ship type, and fuel type.

The objective is to build a production-ready machine learning pipeline and deploy it using an interactive Streamlit web application.

🚀 Live Demo

👉 [Click Here to Try the App](https://huggingface.co/spaces/rshewatkar/ship_co2-emission-predictor)

Deployed on Hugging Face Spaces — No installation required!

🎯 Problem Statement

The shipping industry is a major contributor to global carbon emissions. Predicting CO₂ emissions helps in:

• —🌱 Monitoring environmental impact
• —⚙️ Optimizing fuel efficiency
• —🚢 Supporting sustainable shipping operations

🛠️ Tech Stack

• —Language: Python
• —Libraries: pandas, numpy, scikit-learn, joblib
• —Model: Random Forest Regressor
• —Visualization/UI: Streamlit
• —Deployment: Hugging Face Spaces

⚙️ Features Used

• —Engine Efficiency
• —Fuel Consumption
• —Distance
• —Ship Type
• —Fuel Type

🧠 ML Pipeline

This project uses a scikit-learn Pipeline to ensure robust and consistent predictions:

• —🔹 Data preprocessing using ColumnTransformer
• —StandardScaler for numerical features
• —OneHotEncoder for categorical features

• —🔹 Model training using Random Forest Regressor

• —🔹 Entire pipeline saved using joblib

✅ This approach prevents:

• —Feature mismatch errors
• —Encoding inconsistencies
• —Deployment failures

📊 Model Performance

• —R² Score: 0.9951
• —RMSE: 1012.98

📈 The model explains 99.51% variance, indicating very high prediction accuracy.

🚀 Deployment

The model is deployed using Streamlit on Hugging Face Spaces.

App Features:

• —🎚️ Interactive sliders for numeric inputs
• —📋 Dropdown selection for ship and fuel type
• —⚡ Real-time CO₂ emission prediction

🖥️ How to Run Locally

1️⃣ Clone the repository

git clone <your-repo-link>
cd Ship_Co2_Emission_Prediction

2️⃣ Create virtual environment

conda create -n ml_env python=3.10
conda activate ml_env

3️⃣ Install dependencies

pip install -r requirements.txt

4️⃣ Train the model

python src/train.py

5️⃣ Run the Streamlit app

streamlit run app.py

📁 Project Structure

Ship_Co2_Emission_Prediction/
│
├── app.py                         # Main Streamlit app (UI + model inference)
├── model.pkl                      # Trained ML pipeline (preprocessing + model)
├── requirements.txt               # Python dependencies for reproducibility
├── Dockerfile                     # Container setup for deployment (optional production setup)
├── README.md                      
├── LICENSE                        
│
├── Ship_Co2_Emission_Prediction_project.ipynb   # Jupyter notebook for EDA & experimentation
│
├── data/
│   └── ship_fuel_efficiency.csv   # Dataset used for training the model
│
├── src/
│   ├── train.py                   # Script to train ML pipeline and save model.pkl
│   ├── preprocess.py              # Preprocessing logic (ColumnTransformer setup)
│   │
│   └── __pycache__/               # Python cache files 
│
├── .github/
│   └── workflows/
│       └── sync-to-hub.yml        # GitHub Actions workflow for deployment/automation
│
└── __pycache__/                   # Cache files generated by Python

👤 Author

Rahul Shewatkar