An air quality project that seeks to compare air quality trends between the cities of Nairobi, Mombasa and Kisumu
## Air Quality Monitoring in Kenyan Cities 🏙 🇰🇪
## Project Overview
This project focuses on the **analysis and visualization of air quality data** for three major Kenyan cities: **Nairobi**, **Mombasa**, and **Kisumu**. Using real-time data from the **OpenWeather API**, we explore pollution trends, conduct statistical and machine learning analysis, and create an interactive **Dash** web dashboard.
## Google Colab Notebook Analysis
The project was carried out in a Jupyter Notebook, which was rendered for interactive exploration and analysis. The following key components were analyzed:
1. **Air Quality Analysis**:
- Conversion of timestamps.
- Extraction of pollutant concentrations for different cities.
2. **Data Visualization**:
- Various plots and figures were created using **Matplotlib**, **Plotly**, and **Seaborn**.
- Detailed comparison graphs for pollutants across cities and over time.
---
## Dashboard Visualization (Rendered in Google Colab Notebook)
I built an **interactive dashboard** using `Dash`:
- Select city, month, and date range
- View daily pollutant levels over time
- Compare monthly pollutant averages
- Explore trends and pollutant behavior
> Screenshots of Dashboard:
## Machine Learning Insights
### 1. Classification (AQI Levels)
- Used `main.aqi` to classify air quality as `Good`, `Moderate`, `Unhealthy`, etc.
- Algorithms: Decision Trees, Random Forests
- Input features: pollutant concentrations (`co`, `no2`, `pm2_5`, etc.)
### 2. Clustering Analysis
- Applied **KMeans** and **DBSCAN** to find pollution pattern clusters by time and location
- Helped identify peak pollution hours and typical pollutant profiles per city
### 3. Anomaly Detection
- Used **Multivariate Outlier Detection** and **AUC Score** methods
- Detected unusual spikes in pollutants like `PM₂.₅` and `CO` linked to industrial or traffic events
---
## Technologies Used
- **Google Colab Notebook**: For interactive analysis and visualization
- **Dash / Plotly …