A Comparative Study of Machine Learning Models for Air Quality Index Classification in Urban Bangladesh

This research investigates the applicability of machine learning models to classify and predict air quality index levels for some cities utilizing a data-driven, cost-effective environmental monitoring method. As air pollution poses serious health risks and traditional monitoring systems suffer from limitations, this study explores various machine learning algorithms like Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naive Bayes on pollutants like SO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>, NO, NO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>, CO, O<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</inf>, PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">10</inf> and PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>. Synthetic Minority Oversampling Technique (SMOTE) tackles class imbalance towards air quality groups. This technique enhances the model’s performance, producing more accurate predictions based on skewed data distribution. Models have been run on city-specific datasets and each model’s performance is evaluated by accuracy, confusion matrix and classification reports. Compared to the other models, the Decision Tree model holds the highest accuracy of 99.9%, in Gazipur. This result highlights the superior performance of the Decision Tree model in classifying air quality levels in the city. Although deep learning models (RNN, CNN) were tested, they did not show a notable advantage over traditional machine learning approaches. These findings suggest that machine learning models, enhanced with SMOTE, can improve real-time air quality monitoring, even when dealing with challenges like class imbalance and missing environmental factors. This study paves the way for future research on adaptive, data-driven ecological monitoring systems to support public health and informed policymaking.