Research on Modeling of Air Pollution Prediction Based on Machine Learning
DONG Qianying
LI Huan
MO Xinyue
Abstract：China currently faces the dual challenges of unfavorable meteorological conditions and a significant increase of pol-lutant emissions,making air quality improvement a pressing issue.Accurate prediction of air pollutant concentration is a critical pre-requisite for implementing emergency response measures and improving air quality.Focusing on the prediction of O3 and PM2.5 con-centrations,the data of air pollution and meteorology in Wuhan from 2013 to 2022 are collected in this study.Temporal variation characteristics and influencing factors of the target pollutants are analyzed,and predictors are selected using the Spearman rank cor-relation method.Modeling and prediction are performed based on seven machine learning algorithms respectively,including XG-Boost,AdaBoost and LightGBM.Model performance is evaluated using three statistical metrics,namely MAE,RMSE and R2.Addi-tionally,two ensemble prediction methods,residual correction and entropy weighting methods,are proposed.The experimental re-sults indicate that the temporal variations of O3 and PM2.5 concentrations exhibit"dual-peak"and"U-shaped"patterns,respective-ly,which are influenced by other pollutants and meteorological factors on varying degrees.Among all the individual models,Ada-boost demonstrated the best predictive performance,followed by XGBoost.The ensemble models combining AdaBoost and XGBoost using residual correction and entropy weighting methods are significantly improved in prediction accuracy.
Keywords：air pollution predictionmachine learningcombination predictionresidual correction methodentropy weight-ing method
Publication Date：2025-10-20
Online Publishing Date：2026-01-16(First online date of this platform, not the publication date of the document)
Pages：5( 86-90 )