K-Means Clustering-Based Random Forest System for Air Quality Management and Pollution Reduction in Smart Cities
Abstract
Rapid urbanization has significantly affected ecological sustainability, particularly in emerging nations where cities face elevated air pollution levels. The escalating demands from the WHO revised global air quality recommendations and national legislation have heightened the necessity for efficient urban air quality enhancement measures. However, narrow observations, disconnected citywide operations, lack of standardised processes, insufficient public participation, and poor collaborative governance can impede AQM programs. This study offers a data-driven, smart city paradigm for air quality management and pollution reduction to tackle these issues. Multimodal characteristics are calculated during preprocessing to encapsulate the intricate and diverse dynamics of ambient particulate matter succeeded by the implementation of a time-domain feature selection technique to discern the most pertinent predictors. A thorough analysis is conducted on the performance of various machine learning models, including K-Means, RF, SVM and a hybrid K-Means-RF method. Experimental results demonstrate that the suggested hybrid K-Means–RF multivariate model surpasses traditional methods, with a superior coefficient of determination $\left(\mathbf{R}^{\mathbf{2}}\right)$ of 0.968, thereby exhibiting improved predictive capability. These results highlight the possibility of smart city frameworks integrating analytics to improve air quality monitoring, enable informed decision-making, and execute successful pollution reduction strategies for healthier and more sustainable cities.