Air passenger travel is of great significance to the construction and operation of civil aviation airports. This paper studies the prediction of air passenger index, the main research work is as follows: Firstly, the air passenger index is defined, and the air passenger index is classified using the K-means clustering method. Then, based on the principle of mutual information and correlation, the key influencing factors of air passenger index are selected. This work presents a mutual information-support vector regression(MI-SVR) machine learning model based on mutual information between the key influencing factors and the air passenger index, which is used to predict the air passenger index. Finally, the model is validated by passengers throughput data of the Shanghai airport. The experimental results show that the MI-SVR model is feasible and effective, and compared with classical models, the IM-SVR model has better prediction effect. In addition, it is also showed that the prediction effect of each model is better after introducing influence factors based on mutual information. Overall, the study is helpful to the construction and operation of airports, and it can also help people choose the time to travel by air.