CLUSTERING AND ANOMALY DETECTION OF AIRCRAFT TRAJECTORIES FOR AIRSPACE BEHAVIOR CHARACTERIZATION

Abstract

Background: The augmenting complexity of airspace demands analytical tools that can detect significant patterns of aircraft movement, without resorting to high-dimensional flight data. Even though there has been a boost in the trajectory-based analysis, there are still no integrated frameworks of pattern identification and anomaly detection with minimal positional information. Aim: This study aimed to characterize airspace behavior by clustering aircraft trajectories into distinct movement regimes and detecting anomalous trajectories using a minimal-feature, data-driven approach. Methodology: A quantitative secondary-data design was adopted using time-stamped aircraft trajectory records. After preprocessing and trajectory construction, key kinematic and geometric features were derived, including total distance, mean speed, duration, displacement, and path efficiency. K-means clustering was applied to identify dominant behavioral regimes, while Isolation Forest was used to detect anomalous trajectories. Results: The analysis produced 13,376 valid trajectories from 2,317 aircraft. Four distinct trajectory clusters were identified, showing clear differences in speed, distance, efficiency, and duration. A dominant cluster represented high-speed, long-distance, and highly efficient movement, whereas another captured low-speed, low-efficiency, and irregular trajectories. Anomaly detection identified 669 anomalous trajectories (5% of the dataset), characterized by low displacement, reduced efficiency, longer duration, and lower speeds, indicating fragmented or inefficient movement patterns. Conclusion: These findings demonstrate that meaningful airspace behavior can be extracted from minimal positional data when clustering and anomaly detection are integrated within a unified analytical framework, supporting scalable applications in airspace monitoring and trajectory-based analysis.