SkyTraceX: A Real-Time Short-Horizon Aircraft Trajectory Prediction System Using Gradient Boosted Telemetry Models

Abstract

Accurate short-horizon aircraft trajectory prediction plays an important role in aviation visualization systems, anomaly detection pipelines, and real-time flight intelligence applications. This paper presents SkyTraceX, a lightweight machine learning framework designed to predict aircraft spatial coordinates up to 60 seconds ahead using structured ADS-B telemetry data.

The proposed system utilizes a gradient boosted regression model trained on motion continuity features derived from sequential telemetry observations. The architecture integrates sliding-window feature extraction with Redis-based inference caching and PostgreSQL telemetry storage to support efficient near-real-time deployment.

Experimental evaluation using publicly available ADS-B telemetry datasets demonstrates that the proposed approach improves positional prediction accuracy compared to constant-velocity extrapolation baselines while maintaining low-latency inference suitable for aviation analytics environments.