With the rapid development of China's civil aviation, the air traffic flow in terminal areas is experiencing a consistent and significant increase. The accurate forecast of short-term air traffic flow is of great significance for the efficient implementation of air traffic flow management. To enhance the accuracy of short-term air traffic flow forecast, this paper proposed a model combining empirical mode decomposition (EMD) and long short-term memory (LSTM) based on data differential processing. Firstly, the model performed empirical mode decomposition on short-term air traffic flow sequences. Secondly, to improve prediction accuracy, data difference was utilized to stabilize the time series. Finally, the processed sequences were input into the LSTM network model for prediction, and the final short-term traffic prediction value was obtained through data reconstruction. Experimental verification was conducted using the data from Zhengzhou Xinzheng International Airport. The results demonstrate that the model achieves a significant improvement in prediction accuracy, as indicated by the typical indexes RSME, MAE, and R2, which are 0.29, 0.08, and 96.40%, respectively. This approach outperforms other methods and provides valuable reference for short-term air traffic flow prediction.