Abstract: Based on the air quality data of Zhaotong City in Yunnan Province in the past three years, this paper establishes multiple
linear regression, time series, random forest, and regression tree prediction models, and then selects random forest and regression tree models with higher prediction accuracy to construct a combined model of random forest and regression tree, which uses the standard deviation method to determine the weights. Through the prediction and analysis of the air quality of Zhaotong City in August 2020, it is found that the combined model has higher a prediction accuracy and lower error rate than the single model. Therefore, the model can be widely used in air quality prediction.

Key words: air quality index, random forest, regression tree, combined model, prediction