Sustainability, Journal Year: 2025, Volume and Issue: 17(7), P. 2933 - 2933
Published: March 26, 2025
This study presents an improved deep-learning model, termed Enhanced Time Series Mixer (E-TSMixer), for the prediction of particulate matter. By analyzing temporal evolution PM2.5 concentrations from multivariate monitoring data, model demonstrates significant capabilities while maintaining consistency with observed pollutant transport characteristics in urban boundary layer. In E-TSMixer, a fully connected output layer is proposed to enhance predictive capability complex spatiotemporal dependencies. The relevant data on air quality and traffic flow are fused achieve high-precision predictions through time-series forecasting model. An asymmetric penalty mechanism added dynamically optimize loss function. Experimental results indicate that E-TSMixer achieves higher accuracy PM2.5, which significantly outperforms traditional models. Additionally, intelligent dual regulation fixed dynamic threshold introduced combined decision-making real-time adjustments frequency, routes, timing water truck operation practice.
Language: Английский