Research on Atmospheric Pollutant Concentration Prediction Model Based on the BO-GRU Neural Network with MIV Algorithm

The prediction of atmospheric pollutant concentrations is of crucial importance for environmental protection and safeguarding public health. However, there are some problems with the commonly used air quality prediction methods in China at present, including high model complexity, low prediction accuracy, insufficient spatiotemporal correlation, and poor real-time performance. To enhance the prediction accuracy and real-time capability of atmospheric pollutant concentration forecasting, this study employs the Mean Impact Value (MIV) algorithm for feature selection to determine input variables. Subsequently, a Bayesian Optimization Gated Recurrent Unit (BO-GRU) neural network prediction model is designed. Leveraging the advantages of GRU neural networks, this model fully explores the correlations between various meteorological factors and atmospheric pollutants. Moreover, an intelligent tuning mechanism based on Bayesian optimization is devised to optimize model parameters and structure intelligently, thereby improving the model's learning ability, generalization, and accuracy. Comparative analysis with similar models validates the effectiveness of the proposed model.