Eazi-T/Carbon_Emission_Prediction

Nigeria's Carbon Crossroads: Predicting CO₂ Emissions with Machine Learning # Nigeria's Carbon Crossroads: Predicting CO₂ Emissions with Machine Learning ## Motivation Nigeria is urbanizing rapidly, its economy is growing, and energy demand is climbing. But what is actually driving its CO₂ emissions — and can socioeconomic indicators help us predict and understand future trajectories? This project applies the CRISP-DM data science process to three decades of World Bank data (1990–2023) to answer that question. **Questions of Interest:** - What factors drive Nigeria's CO₂ emissions the most? - Can we accurately predict future emissions from socioeconomic indicators? - How has rapid urbanization affected emissions over time? - What would a green transition look like in measurable terms? --- ## Libraries Used | Library | Purpose | |---|---| | `pandas` | Data loading, cleaning, and manipulation | | `numpy` | Numerical operations and log transformation | | `matplotlib` | Visualizations and plots | | `scikit-learn` | Ridge Regression, cross-validation, scaling | | `xgboost` | XGBoost model training and feature importance | | `shap` | SHAP values for XGBoost model explainability | --- ## Repository Structure | File | Description | |---|---| | `CO2_Emission_Prediction.ipynb` | Main Jupyter notebook with full analysis, comments, and visualizations | | `Co2_data.xlsx` | Raw dataset (World Bank, 1990–2023) containing CO₂, GDP per capita, energy use, renewable share, industry %, urban population %, and population growth | | `README.md` | This file | --- ## Summary of Results ### Exploratory Data Analysis - CO₂ emissions nearly doubled from ~75 Mt in 1990 to ~126 Mt in 2023 - GDP per capita is heavily right-skewed and was log-transformed before modeling - Industry's share of GDP *fell* from ~37% to ~18% over the period — a counterintuitive finding given rising emissions - Renewable energy share and CO₂ showed the strongest correlation (r = −0.94) - Year and Urban population were found to be perfectly collinear (r = 1.00); Year was dropped # …