SFFO-RL-intelligent environmental monitoring: smart sensing networks with artificial intelligence

Rapid urbanization has led to increasingly severe environmental degradation, necessitating the development of intelligent, adaptable, and explainable monitoring systems for sustainable urban management. Existing modeling frameworks are limited by problems related to sensor noise, data inconsistencies, poor adaptability, and a lack of explanation of how models arrive at their predictions. This study proposes a Satin Firefighter Optimization-Reinforcement Learning (SFFO-RL) model within an Intelligent Environmental Monitoring Framework (IEMF) to enable accurate, interpretable environmental assessments. The IEMF employs adaptive filtering for sensor noise, an Isolation Forest outlier-removal method, and Min-Max normalization of the data, followed by Multi-order Neighbor Feature Fusion (MNFF) for comprehensive feature representation. The SFFO-RL contributes to more stable convergence in environmental monitoring, a better exploration-exploitation balance, and improved adaptability in predicting pollutant concentrations when one or more pollutants change over time. Predictions were applied to a novel Environmental Health Index (EHI) to provide early warning alerts, while model visualization explained how data were transformed into predictions. Ultimately, the proposed SFFO-RL achieved 97.1% accuracy, 97.8% F1 score, 97.6% recall, and 98.21% precision.