Fig. 2

The study explored the predictive value of variables for OSAHS using machine learning. a The importance ranking of variables in KNN model. b The importance ranking of variables in RF model. On the left was the selection of random decision trees, with the ordinate representing the error rate, and the optimal tree corresponded to the lowest error rate; in the right figure, the abscissa was the importance value and the ordinate was the variable. c The bar chart of variable importance in the SHAP model. The abscissa was the shap value and the ordinate was the mean value of variables. d The variable dependence graph of pollinosis. The abscissa was pollinosis (1 and 2) and the ordinate was the distribution of shap values