🌍 Global PM2.5 Next-Day Forecasting

Predicting next-day fine particulate matter (PM₂.₅) concentrations using Machine Learning techniques.

This repository contains code, data workflow, and visualizations from the Kaggle-based notebook for global air quality prediction.
It leverages meteorological and environmental datasets to build high-performance models that forecast PM₂.₅ concentration levels across multiple regions worldwide.

---

🚀 Project Overview

Air pollution remains one of the most critical global challenges.
This project develops a data-driven machine learning framework for next-day PM₂.₅ forecasting using publicly available datasets.

Key Features:

• Data preprocessing and feature engineering from multi-source air quality datasets
• Model training using tree-based ensemble methods (e.g., XGBoost, LightGBM)
• Performance evaluation with RMSE and R² metrics
• SHAP interpretability for feature impact analysis
• Visual dashboards and plots for spatial-temporal understanding

---

🧠 Methodology

1. Data Aggregation:
   Merging global PM₂.₅ datasets with weather parameters such as temperature, wind speed, humidity, and pressure.

2. Preprocessing & Cleaning:
   Handling missing values, scaling features, and temporal alignment.

3. Model Development:
   Training machine learning regressors like:

   • XGBoost
   • Random Forest
   • LightGBM
   • Linear Regression

4. Evaluation:

   • RMSE, MAE, and R²
   • SHAP-based feature importance visualization

5. Forecast Generation:
   Produces next-day PM₂.₅ predictions for multiple global locations.

---

📊 Results & Visualization

You can explore the interactive visualizations and full notebook here:
🔗 View the Project Dashboard

---

🧩 Repository Structure

│
├── data/                     # Raw and processed datasets (not pushed due to size)
├── docs/                     # HTML outputs for GitHub Pages
│   ├── index.html
│   └── notebook.html
│
├── global-analysis-next-day-pm2-5-ml.ipynb   # Kaggle notebook
├── requirements.txt           # Environment dependencies
└── README.md                  # Project documentation