A novel combined multi-variate prediction framework for air pollution based on feature selection and deep learning models

Air pollution forecasting offers crucial support for pollution control and regulation. However, the non-linearity, volatility, and complexity of air pollution data pose significant challenges to accurate prediction. To address these challenges, this study develops a multi-variate combined forecasting framework based on feature selection methods, linear and nonlinear forecasting models, and intelligent optimization algorithms. Firstly, the framework analyzes the impact of each subset after feature selection on model performance. Secondly, multiple benchmark models based on different feature subsets and various prediction models are constructed as comparison models to further test the performance of the proposed prediction framework. Finally, two real-world datasets, statistical hypothesis testing, and several evaluation metrics are used to validate the prediction performance of the proposed framework. The Root Mean Square Error (RMSE) of the proposed model on the Site 1 dataset is 1.9762, representing an improvement of approximately 4 % compared to the best-performing benchmark model. Statistical testing further confirms that the proposed model significantly outperforms the benchmark model. The proposed combined multi-variate forecasting framework provides a heuristic framework for building combined prediction models with high prediction accuracy and robustness.