IoT-Based Transfer Learning and Genetic Algorithm Approaches for Smart Environmental Pollution Detection and Management

The identification of environmental pollution remains to be a significant issue, as conventional technologies lack scalability and flexibility to dynamic, real-time situations. This study presents an IoT-enabled hybrid model that combines transfer learning and genetic algorithms to overcome these restrictions. IoT networks enable real-time, extensive data acquisition, whereas transfer learning uses pre-trained models for effective feature extraction and domain adaption. Genetic algorithms enhance hyperparameters and sensor configurations, guaranteeing energy economy and optimal model performance. The suggested CNN-LSTM architecture utilising multi-scale attention attains a training accuracy of 95.6% and an F1 score of 0.96, exceeding the performance of conventional Random Forest and CNN models in accuracy, recall, and precision. The comparative study confirms the model's strength and flexibility, illustrating its capacity for scaled, practical applications. This framework creates a dependable and effective system for intelligent environmental pollution control, providing substantial improvements in detection accuracy and operational efficacy compared to current methods.