Research On Air Quality Prediction Model Based On Spatio-Temporal Features

Ecological civilization construction is a millennium-long endeavor that concerns the sustainable development of the Chinese nation.It is essential to plan economic and social development from the perspective of harmonious coexistence between humans and nature.With the rapid progress of urbanization and industrialization,air pollution has become one of the significant environmental challenges worldwide.The Air Quality Index(AQI)is an important indicator for assessing the level of air pollution,and accurate AQI predictions can assist countries,businesses,and individuals in formulating timely plans.Existing AQI prediction models primarily rely on temporal features.The main focus of this study is how to construct an air quality prediction model that simultaneously incorporates both temporal and spatial features to enhance prediction accuracy.This article begins by exploring the composition of air pollutants,methods for assessing air quality,and factors influencing the AQI.It then describes the existing AQI prediction models.Next,by constructing an adjacency matrix based on the geographic locations of different districts in Beijing,spatial features are aggregated using GCN.Subsequently,GBiLSTM,which combines meteorological and air pollutant temporal monitoring data with a two-layer GCN and BiLSTM,is designed to incorporate both temporal and spatial features.Finally,an attention mechanism is introduced to create the AGBiLSTM for air quality prediction.In this study,five baseline models including BiLSTM,LSTM,GRU,GCN,and GBiLSTM are compared with the AGBiLSTM model through experimental analysis.A dataset consisting of 35,064 rows of temporal monitoring data related to air pollutants and meteorology in Beijing is collected,and the performance of each prediction model is analyzed based on Mean Absolute Error(MAE)and Root Mean Square Error(RMSE).The experimental results demonstrate that the AGBiLSTM prediction model achieves the best performance in terms of both MAE and RMSE,indicating its high accuracy and predictive capability in air quality forecasting.By incorporating the adjacency matrix,GCN,BiLSTM,and attention mechanism,a more accurate air quality prediction model that combines temporal and spatial features is constructed.However,due to the complexity of atmospheric environmental changes,the AGBiLSTM air quality prediction model still has certain limitations,and there is room for further optimization in multiple dimensions.