Hybrid stacked bidirectional deep learning with attention based air pollution prediction for environmental sustainability

Environmental sustainability for air pollution monitoring necessitates utilization of advanced technologies like internet of things (IoT) devices and Artificial Intelligence (AI) AI to track air quality in real-time, controlling policy and measures to reduce to mitigate pollution for ensuring long-term environmental health. This mechanism aids sustainable development by safeguarding general public, combating climate change and establishing more flexible communities. Hence pollutant level monitoring is of paramount significance to maintain the pollutant concentrations under control. Nevertheless, pollutant concentration monitoring is not easy as it necessitates installing monitoring stations to obtain pertinent pollutant data that comes with high installation and maintenance costs. In this research, an attempt has been made to simulate the concentrations at three districts in Dhaka, Bangladesh using Deep Learning (DL) methods by applying Hybrid Stacked Bi-directional Long Short-Term Memory with Attention Mechanism (HSBi-LSTMA) method for environmental sustainability. An efficient model specifically tailored for encoding and generating multi-class imbalance results using Label Encoded Synthetic Edited Nearest Neighbor-based pre-processing for different air pollution concentration features provided as input is introduced. Next, Hybrid Stacked Bi-directional Long Short-Term Memory with Attention Mechanism first used attention mechanism to select the most influential features of the original air pollution concentration features and then used Hybrid Stacked Bi-directional Long Short-Term Memory to generate the classified results for air pollution prediction. Extensive experiments were conducted on Air Pollution Monitoring Generated Using IoT dataset, comparing the HSBi-LSTMA method with two existing methods. The results demonstrate that the HSBi-LSTMA method significantly improves the air pollution prediction accuracy by 25% and reduces the mean absolute error by 46% as compared to the existing methods. With this the HSBi-LSTMA method represents a notable evolution in advancing environmental sustainability.