In order to prevent air pollution and to predict urban pollution grade and factors which play a leading role in air pollution based on each index value of air pollution， air quality classification prediction method based on random forest is proposed. Random forest model directly gives the scores influencing the importance of air quality indicators so as to find the most important influencing factor. Comparison of different data mining methods shows that the accuracy of random forest classification prediction is the highest， therefore， this model can be widely used in air quality prediction. Testing results indicate that random forest method is not easy to be affected by noise and has low generalization error. The importance scores show that fine particulate and inhaled particulate are the most important two factors. By taking Baoding City as an example， the suggestions for raising air quality are provided accordingly.