Asiboje-efe/AirQo-African-Air-Quality-Prediction-
Domaine:
environment and energyclimateType de record:
model
Créateur:
Asi
Hôte:
Air quality prediction in Africa using Light Gradient Boost Model and Random Forest Regressor
# AirQo-African-Air-Quality-Prediction-
Background
• Overview
• Project question
• Evaluation metrics
Findings
Conclusion
## Background
### Overview
AirQo African Air Quality Prediction aims to predict air quality around Africa using Sentinel 5P satellite data. The project focuses on estimating pm2_5 levels from satellite observations based on Aerosol Optical Depth (AOD) for four cities in four African countries. Air pollution is a major environmental health risk globally, it contributes to 7 million premature deaths globally each year with developing countries being the most affected World Health Organization (2021).
### Project Question
Can Sentinel 5P data be used to predict air quality around Africa for environmental justice?
### Evaluation Metrics
The performance of the air quality prediction models will be evaluated using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and R-squared.
## Findings
• The dataset contains 8,071 rows and 80 columns. It is multi-dimensional dataset with numerous environmental variables.
• The data covers 4 countries: Nigeria, Kenya, Burundi, and Uganda. It includes 4 cities: Lagos, Nairobi, Bujumbura, and Kampala. This indicates a focus on urban air quality in East and West Africa.
• PM2.5 is the main air quality indicator, with levels ranging from 1.2 to 456.19 μg/m³. Other pollutants measured include sulphur dioxide, carbon monoxide, nitrogen dioxide, formaldehyde, and ozone.
#### Target variable (pm2_5) has values skewed to the right
#### The target variable (pm2_5) has some outliers
#### Correlation heatmap of target varaible (pm2_5) and other variables
There are weak positive and negative associations between pm2_5 and other variables
#### Original distribution of pm2_5 and log transformed pm2_5
The original distribution of PM2.5 is heavily right-skewed, indicating a long tail of high values. The log transformation effectively reduces the skewness in the data, making the distribution more bell-shaped.
K …