DP-CDUN-GTO: An Enhanced Network Approach with Giant Trevally Optimizer for Accurate Air Quality Prediction

Air pollution is a serious problem in smart cities that damages the environment and people's health. Predicting air pollution can provide pertinent parties the knowledge they need to act appropriately. Notwithstanding the importance of this field, there are problems with the current research, including low accuracy and inaccurate real-time updates. In order to optimize performance, this study offers a novel hybrid model called DP-CDUN-GTO, which is based on the Giant Trevally Optimizer and a Dynamic Path-Controllable Deep Unfolding Network(DP-CDUN). The suggested model uses a DP-CDUN layer to extract a feature set from the training dataset and uses the same network to accomplish sequencing learning. Furthermore, the model builds a novel motivating model by applying a Curiosity-based Motivation method, which improves the reconstruction of the deep unfolding network. The DP-CDUN weights are optimized by the Giant Trevally Optimizer in order to reduce prediction error. For comparison, the authors use an online dataset from Salt Lake City, USA's Air Pollution Monitoring, which includes five air quality indicators: SO2, CO, O3, NO2. The suggested attains R2-Score = 0.1124, Mean Absolute Percentage Error = 0.0784, and Mean Absolute Error = 0.0052.