Application of machine learning to predict respiratory disease deaths using meteorological and air pollution factors: a case study of Haidian District, Beijing

With the development of urbanization and industrialization, increasingly severe air pollution and frequent extreme weather events pose a significant threat to public health. The objective of this study was to evaluate the effects of meteorological factors and air pollution on respiratory deaths. The study dataset includes meteorological, air pollutant, and respiratory disease death data from Haidian District, Beijing, China, from January 1, 2014 to July 31, 2024. A random forest (RF) model was used to analyze the effects of meteorological factors and air pollutant levels on respiratory disease mortality, and the factors influencing respiratory disease mortality were analyzed in combination with SHapley Additive exPlanations (SHAP). The results of Spearman correlation analysis and the RF model showed that SO₂, NO₂, PM_2.5 and PM₁₀ concentrations were positively correlated with respiratory mortality, whereas the minimum temperature was negatively correlated with respiratory mortality. In addition, the model showed better prediction performance in winter than in other seasons. Furthermore, the SHAP global feature results indicate that the minimum temperature is the most significant factor affecting mortality from respiratory diseases.The results show that the RF model has the potential to predict deaths from respiratory diseases, since it can effectively combine meteorological and air pollution data. Combined with SHAP, it can further enhance the interpretability of the machine learning model. This study should provide strong support for policymakers in scientifically formulating targeted air quality control measures, health warnings about extreme temperatures, and prevention and control strategies for seasonal respiratory diseases.