Scientific and accurate prediction of short-term airport traffic flow will provide essential and timely information for decision-making and planning. This paper proposes a short-term airport traffic flow prediction model based on Long-Short Term Memory Recurrent Neural Network. Extracting hourly airport traffic flow data of Xi'an XianYang Internation Airport in 2014, we construct LSTM neural network prediction model with compressed auto encoder to predict airport traffic flow in a future day. The evaluation was performed with Monte-Carlo cross validation. Considering the season factor of airport traffic flow, the results were illustrated on both high season and low season. Compared with other comparative machine learning approaches, LSTM NN prediction model outperforms with regard to Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), which proves that LSTM model is effective for airport traffic flow prediction.