China major cities air quality index forecasting using integrative machine learning models: A perception for sustainable cities

AbstractAccurate air quality forecasting is essential process for establishing sustainable cities and healthy communities. The present study introduces an innovative integrative machine learning (ML) model to overcome the limitations of traditional ML models and prior studies in capturing the complicated nonlinearity of air quality index (AQI). The integration of weighted least squares support vector machine with generalized regression model (WLSSVM-GRM) was adopted as predictive model for AQI. An optimized multivariate variational mode decomposition (OMVMD) method was introduced to decompose the input variables and increase the forecasting accuracy. Further, the proposed model uses an efficient light gradient boosting machine (LGBM) model enhanced by the self-adaptive teaching learning-based with differential evolution (SATLDE) algorithm to identify the most significant features influencing AQI. The proposed OMVMD-WLSSVM-GRM model showed an outstanding performance in forecasting AQI across Beijing, Wuhan, and Xi’an in China with correlation coefficient (R ≈ 0.99, 0.97, and 0.98), outperforming other benchmark models. Also, the proposed model attained low extreme errors ((Formula presented) 97.3, 78.3 and 78.3), and better vicis symmetric distance stability (VSD ≈ 342.6, 632.3, and 638.8). Uncertainty analysis confirmed the proposed model reliability, with smaller forecasting intervals and narrower uncertainty ranges. Overall, the developed model provided reliable air quality forecasting and possibly enhanced environmental management and preserved the health of the community.