Virtual Booth

Air pollution is one of the major environmental problems that is affecting public health 
and urban sustainability. Traditional air quality assessment procedures may have 
difficulty understanding complex environmental data and providing meaningful 
insights to make decisions. Therefore, this project will focus on developing an 
integrated air quality analysis and classification system with machine learning and 
Rule-Based AQI assessment to improve the evaluation of air quality conditions. The 
main goal of this project is to develop a MATLAB analytical system to analyze air 
quality data, classify pollution levels, identify abnormal pollution patterns and present 
analysis results on an interactive dashboard. The proposed system uses air quality 
variables such as particulate matter (PM2.5 and PM10), gases, temperature, and 
humidity to assess the air quality system performance. We use two datasets: Dataset 1 
(Kaggle) and Dataset 2 (Zenodo) to compare the performance of different datasets. 
The system uses supervised machine learning classifiers such as Random Forest and 
Gradient Boosting for air quality classification. Based on validation performance, 
automatic machine learning selection is introduced to select the best classifier. We also 
use Rule-Based AQI for AQI values, air quality categories, dominant pollutants and 
health advice. Other analysis techniques like Z-score anomaly detection and 
environmental comparison analysis are incorporated to further analyze pollution 
patterns. The outputs are visualized by MATLAB tools as well as on a browser-based 
dashboard to facilitate interpretation and user interaction. The testing results show that 
the system handles dataset processing, machine learning evaluation, automatic model 
selection, AQI assessment, anomaly detection and visualization. The automatic model 
selection approach successfully identifies the most suitable classifier according to 
dataset characteristics, where Random Forest achieved the best overall performance 
for Dataset 1, while Gradient Boosting was selected for Dataset 2. In summary, this 
project demonstrates the effectiveness of combining machine learning classification, 
Rule-Based AQI assessment, anomaly detection, and dashboard visualization into a 
single air quality analysis platform.