Clustering and BiLSTM Network for Aircraft Trajectory Prediction Model

Abstract

The increasing demand for air travel requires the development of more accurate aircraft trajectory prediction methods to optimize airspace utilization and enhance safety. This paper presents a hybrid approach for single-flight-route trajectory prediction that employs the K-means clustering and Bidirectional Long Short-Term Memory (BiLSTM) networks. The primary objective is to develop a deep learning model that effectively predicts aircraft trajectories. Additionally, this research investigates the influence of trajectory clustering on prediction accuracy. To fulfill the objectives, a four-step methodology: data preprocessing, model construction, validation testing, and analysis is employed. Real-world historical flight data is used to train the BiLSTM model after being clustered with K-means. The model's performance is evaluated using randomized enroute flight data and various metrics like mean squared error and root mean squared error. This research is successful in accurately predicting the flight and the clustering process was proven to increase prediction accuracy by 15 percent in latitude, and 10 percent in longitude.