Logo Lanfrica

nguoimay1103/Predict-CO2-Emissions-in-Rwanda

Domain:

climateenvironment and energy

Record type:

project
Creator:
ngu
Host:
# 🌍 CO2 Emission Modeling in Rwanda This repository contains a **machine learning project** for predicting COβ‚‚ emissions in Rwanda. It includes: - **Jupyter Notebook** for exploratory data analysis (EDA) and model training. - **Streamlit Dashboard** for interactive visualization of COβ‚‚ emissions on the map of Rwanda. - **PDF report** summarizing results and insights. ## πŸ“Š Project Overview - **Dataset**: Kaggle Playground Series S3E20 - **Goal**: Predict COβ‚‚ emissions based on weekly measurements across locations in Rwanda. - **Users**: Researchers & environmental policy managers. - **Machine Learning Models Used**: - XGBoost - CatBoost - LightGBM - Random Forest - **Feature Engineering**: - Rolling means (moving averages). - Sine/cosine transformations for cyclical features (weeks of year). - Normalization & scaling. - Handling missing values & outliers. - **Evaluation Metric**: `RMSE` (Root Mean Squared Error). ## βš™οΈ How to Run 1. Install dependencies: `pip install -r requirements.txt` 2. Run the notebook: `jupyter notebook co2-modeling\ (9).ipynb` ## πŸ“¦ Dependencies **Main Python libraries used:** - pandas, numpy – data processing - geopandas, folium, pydeck, plotly – geospatial visualization - scikit-learn – ML utilities - xgboost, catboost, lightgbm, randomforest – ML models - optuna – hyperparameter tuning (optional) - streamlit – dashboard UI ## πŸ“‘ Results Best model achieved low RMSE for emission prediction. Clear spatial and temporal patterns in Rwanda’s COβ‚‚ emissions. Dashboard enables easy exploration by policymakers and researchers. ## πŸ“œ License This project is released under the MIT License. Feel free to use, modify, and share with attribution. ## Demo