Aircraft running off the runway during landing is a high risk of accident. Due to the difficulty of circumvention and the frequency of accidents, a neural network-based online prediction method is proposed to predict the landing distance of the aircraft in advance. Specifically, the wind field conditions can be adapted by the method to realize dynamic prediction. The B737-800 reference object was selected as the desktop simulation platform for the aircraft landing process. Multiple wind field conditions and flight trajectory data are simulated and collected, and key feature data are preprocessed, and a long and short-term memory (LSTM) neural network is selected to build the prediction model. The results indicate that the prediction error for longitudinal landing distance is 188.4 m and for lateral landing deviation distance is 43.5 m. For runways that are at least 2000 m long and 40 m wide, the prediction errors can provide the accuracy required for pilots to make go-around decisions in the air.