Anomaly Detection in Aircraft Trajectories using Machine Learning

The aviation industry is rapidly embracing data-driven techniques to enhance safety, efficiency, and situational awareness. Anomaly detection in aircraft trajectories plays a crucial role in identifying irregular flight patterns that may indicate safety risks, system malfunctions, or security threats. This paper presents a comprehensive study of various machine learning (ML) approaches for anomaly detection in aircraft trajectories using Automatic Dependent Surveillance-Broadcast (ADS-B) data. We explore both supervised and unsupervised models, including Isolation Forests, Autoencoders, and LSTM-based sequence models, to detect outliers in spatiotemporal data. We evaluate the models using a real-world ADS-B dataset and compare their performance using precision, recall, and F1score metrics. Our findings suggest that hybrid deep learning models outperform classical methods in complex trajectory anomaly detection.