🌍 Global Air Quality Dataset Analysis

CSC380 - Introduction to Data Science | Semester Project
University of Engineering and Technology, Lahore | Fall 2025

A comprehensive data science project analyzing global air quality measurements from major cities worldwide, featuring exploratory data analysis, machine learning model building, and actionable environmental recommendations.

---

📋 Table of Contents

• Overview
• Dataset Description
• Project Structure
• Features
• Installation
• Usage
• Analysis Pipeline
• Machine Learning Models
• Results
• Visualizations
• Key Findings
• Technologies Used
• Contributing
• License

---

🎯 Overview

Air pollution is a global concern affecting millions of people worldwide. This project analyzes the Global Air Quality Dataset containing 10,000 records of air quality measurements from various prominent cities worldwide. The analysis includes:

• Data Preprocessing & Cleaning - Handling missing values, outliers, and feature engineering
• Exploratory Data Analysis (EDA) - Comprehensive univariate, bivariate, and correlation analysis
• Machine Learning - Building and comparing classification models for AQI prediction
• Model Evaluation - Performance comparison with accuracy, precision, recall, and F1-score
• Recommendations - Strategies for environmental improvement

🎯 Objectives

1. Analyze air quality trends and patterns across different cities and countries
2. Build predictive models for AQI (Air Quality Index) categories
3. Understand the impact of pollution on health
4. Suggest strategies for environmental improvement

---

📊 Dataset Description

Feature	Description	Data Type
City	Name of the city	Object
Country	Country where city is located	Object
Date	Date of measurement	Object
PM2.5	Fine particulate matter (µg/m³)	Float64
PM10	Coarse particulate matter (µg/m³)	Float64
NO2	Nitrogen dioxide (µg/m³)	Float64
SO2	Sulfur dioxide (µg/m³)	Float64
CO	Carbon monoxide (mg/m³)	Float64
O3	Ozone (µg/m³)	Float64
Temperature	Temperature at measurement time (°C)	Float64
Humidity	Humidity level (%)	Float64
Wind Speed	Wind speed (m/s)	Float64

Dataset Statistics

Feature	Mean	Std Dev	Min	Max
PM2.5	77.45	41.93	5.02	149.98
PM10	104.44	55.06	10.00	200.00
NO2	52.20	27.32	5.01	100.00
SO2	25.34	14.09	1.00	49.99
CO	5.05	2.85	0.10	10.00
O3	106.03	55.08	10.04	200.00
Temperature	14.90	14.44	-10.00	40.00
Humidity	55.08	25.98	10.01	99.99
Wind Speed	10.23	5.63	0.50	20.00

• Total Records: 10,000
• Cities Covered: 20
• Countries Covered: 19
• Date Range: January 1, 2023 to December 28, 2023

---

📁 Project Structure

ids/
├── air_quality_analysis.py      # Main analysis script (971 lines)
├── global_air_quality_data_10000.csv  # Raw dataset
├── cleaned_air_quality_data.csv # Preprocessed dataset
├── model_results.csv            # ML model performance results
├── REPORT_CONTENT.txt           # Detailed project report
├── README.md                    # This file
├── requirements.txt             # Python dependencies
└── figures/                     # Generated visualizations
    ├── 01_univariate_pollutants.png
    ├── 02_univariate_meteorological.png
    ├── 03_aqi_distribution.png
    ├── 04_city_country_distribution.png
    ├── 05_bivariate_pollutants_aqi.png
    ├── 06_bivariate_meteo_aqi.png
    ├── 07_boxplot_by_category.png
    ├── 08_correlation_matrix.png
    ├── 09_aqi_by_city.png
    ├── 10_aqi_by_country.png
    ├── 11_temporal_analysis.png
    ├── 12_monthly_pollutant_trends.png
    ├── 13_seasonal_city_comparison.png
    ├── 14_pairplot.png
    ├── 15_confusion_matrices.png
    ├── 16_model_comparison.png
    ├── 17_feature_importance.png
    └── 18_cv_scores.png

---

✨ Features

Data Preprocessing

• ✅ Missing value detection and handling (interpolation & mode imputation)
• ✅ Outlier detection using IQR method
• ✅ Outlier handling using Winsorization
• ✅ Date parsing and temporal feature extraction (Year, Month, Day, Season, Quarter)
• ✅ AQI calculation based on US EPA standards
• ✅ Feature scaling (StandardScaler & MinMaxScaler)
• ✅ Label encoding for categorical target variable

Exploratory Data Analysis

• 📊 Univariate analysis of pollutants and meteorological variables
• 📈 Bivariate analysis with scatter plots and trend lines
• 🔥 Correlation matrix heatmap
• 🏙️ Comparative analysis across cities and countries
• 📅 Temporal/Seasonal trend analysis

Machine Learning

• 🌲 Random Forest Classifier
• 🚀 Gradient Boosting Classifier
• 📈 Logistic Regression
• 🎯 K-Nearest Neighbors (KNN)
• ⚡ Support Vector Machine (SVM)

---

🔧 Installation

Prerequisites

• Python 3.8 or higher
• pip (Python package installer)

Setup

1. Clone the repository

git clone https://github.com/muhammad-rateeb/global-air-quality-analysis.git
cd global-air-quality-analysis

2. Create a virtual environment (recommended)

python -m venv .venv
# Windows
.venv\Scripts\activate
# Linux/Mac
source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

---

🚀 Usage

Run the complete analysis

python air_quality_analysis.py

This will:

1. Load and preprocess the dataset
2. Perform comprehensive EDA
3. Generate all visualizations in the figures/ folder
4. Train and evaluate 5 ML models
5. Save results to model_results.csv
6. Save cleaned data to cleaned_air_quality_data.csv

---

🔄 Analysis Pipeline

graph TD
    A[Raw Data] --> B[Data Loading]
    B --> C[Missing Value Analysis]
    C --> D[Outlier Detection & Handling]
    D --> E[Feature Engineering]
    E --> F[AQI Calculation]
    F --> G[Feature Scaling]
    G --> H[Train-Test Split 80:20]
    H --> I[Model Training]
    I --> J[Model Evaluation]
    J --> K[Results & Recommendations]

Loading

Preprocessing Steps

Step	Method	Description
Missing Values	Interpolation + Mode	Linear interpolation for numerical, mode for categorical
Outliers	IQR Method	Detection using Q1-1.5×IQR to Q3+1.5×IQR
Outlier Handling	Winsorization	Capping values at IQR bounds
Feature Scaling	StandardScaler	Z-score normalization
Train-Test Split	Stratified	80% training, 20% testing

AQI Categories

AQI Range	Category	Health Implication
0 - 50	Good	Air quality is satisfactory
51 - 100	Moderate	Acceptable; moderate concern for some
101 - 200	Unhealthy	Everyone may experience health effects
201+	Hazardous	Health alert; everyone affected

---

🤖 Machine Learning Models

Models Implemented

1. Random Forest Classifier - Ensemble of decision trees with bagging
2. Gradient Boosting Classifier - Sequential ensemble with boosting
3. Logistic Regression - Linear classification with regularization
4. K-Nearest Neighbors - Instance-based learning
5. Support Vector Machine - Kernel-based classification

Hyperparameters

Model	Key Parameters
Random Forest	n_estimators=100, max_depth=None, random_state=42
Gradient Boosting	n_estimators=100, learning_rate=0.1
Logistic Regression	max_iter=1000, multi_class='multinomial'
KNN	n_neighbors=5, metric='minkowski'
SVM	kernel='rbf', C=1.0

---

📈 Results

Model Performance Comparison

Model	Train Accuracy	Test Accuracy	Precision	Recall	F1-Score	CV Mean	CV Std
Random Forest	100%	99.9%	99.85%	99.9%	99.87%	99.89%	0.07%
Gradient Boosting	100%	99.9%	99.85%	99.9%	99.87%	99.93%	0.05%
Logistic Regression	95.61%	96.15%	95.84%	96.15%	95.91%	95.43%	0.41%
KNN	94.0%	89.1%	88.87%	89.1%	88.93%	89.44%	0.49%
SVM	97.51%	97.15%	97.06%	97.15%	97.09%	96.11%	0.39%

🏆 Best Models

• Random Forest and Gradient Boosting achieved the highest accuracy (99.9%)
• Both models showed excellent generalization with minimal overfitting
• Cross-validation confirmed consistent performance

---

📊 Visualizations

The project generates 18 comprehensive visualizations:

Sample Visualizations

📈 Click to see visualization descriptions

Figure	Description
01_univariate_pollutants.png	Distribution histograms of all 6 pollutants
02_univariate_meteorological.png	Distribution of temperature, humidity, wind speed
03_aqi_distribution.png	AQI histogram and category bar chart
04_city_country_distribution.png	Record counts by city and country
05_bivariate_pollutants_aqi.png	Scatter plots: pollutants vs AQI with trend lines
06_bivariate_meteo_aqi.png	Scatter plots: meteorological variables vs AQI
07_boxplot_by_category.png	Box plots of pollutants by AQI category
08_correlation_matrix.png	Heatmap of feature correlations
09_aqi_by_city.png	Average AQI comparison across cities
10_aqi_by_country.png	Average AQI comparison across countries
11_temporal_analysis.png	Monthly AQI trends over time
12_monthly_pollutant_trends.png	Monthly pollutant concentration patterns
13_seasonal_city_comparison.png	Seasonal AQI variations by city
14_pairplot.png	Pairwise relationships between features
15_confusion_matrices.png	Confusion matrices for all 5 models
16_model_comparison.png	Performance metrics comparison chart
17_feature_importance.png	Random Forest feature importance
18_cv_scores.png	Cross-validation score comparison

---

🔍 Key Findings

Pollutant Analysis

• PM2.5 shows the strongest correlation with AQI
• Ozone (O3) levels vary significantly with temperature
• CO concentrations are relatively stable across regions

Geographic Insights

• Major industrial cities show consistently higher AQI values
• Coastal cities benefit from better air dispersion
• Seasonal patterns differ between hemispheres

Model Insights

• Ensemble methods (Random Forest, Gradient Boosting) significantly outperform other models
• Feature importance reveals PM2.5, PM10, and O3 as top predictors
• All models show good generalization with stratified sampling

Health Recommendations

1. 🚗 Reduce vehicle emissions in urban areas
2. 🏭 Implement stricter industrial emission controls
3. 🌳 Increase green spaces and urban forests
4. 📊 Establish real-time air quality monitoring systems
5. 🎓 Public awareness campaigns on air quality impacts

---

🛠️ Technologies Used

Category	Technologies
Language	Python 3.x
Data Processing	Pandas, NumPy
Visualization	Matplotlib, Seaborn
Machine Learning	Scikit-learn
IDE	VS Code / Jupyter Notebook

Python Libraries

pandas          # Data manipulation and analysis
numpy           # Numerical computing
matplotlib      # Static visualizations
seaborn         # Statistical visualizations
scikit-learn    # Machine learning algorithms

---

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

👨‍💻 Author

Muhammad Rateeb

• GitHub: @muhammad-rateeb
• University: University of Engineering and Technology, Lahore
• Course: CSC380 - Introduction to Data Science

---

🙏 Acknowledgments

• Dataset: Kaggle - Global Air Quality Dataset
• University of Engineering and Technology, Lahore
• CSC380 Course Instructors

---

Made with ❤️ for Data Science