A Continuous Flight Arrival Time Prediction Model for Air Cargo Based on ADS-B Trajectory

Predicting estimated time of arrival (ETA) is of significant importance in air cargo transportation. This study aims to compare the performance of different models in predicting Estimated Time of Arrival (ETA) for flights and propose improvement methods. Firstly, we introduce two prediction methods based on statistical models and LSTM models. For the statistical model, we adopt traditional statistical methods and analyze historical data to predict ETA. As for the LSTM model, we utilize deep learning techniques to enhance the accuracy of flight arrival time prediction. Subsequently, we provide detailed descriptions of the dataset construction and model training process. In the experimental section, we compare the prediction performance of different models under different time spans and extreme scenarios. The results indicate that the prediction accuracy based on the LSTM model is significantly superior to that of traditional statistical models, especially for longer time spans and extreme scenarios. Additionally, we propose improved logic strategies to further enhance the accuracy and robustness of predictions. In summary, this study provides new insights and methods for improving the prediction of flight arrival times, contributing to both theoretical and practical advancements in the field of air transportation.