Graph machine learning for flight delay prediction due to holding manouver

TL;DR

This paper tackles predicting flight delays caused by holding maneuvers by modeling the air traffic network as a directed graph and framing holding prediction as an edge‑prediction task. It compares CatBoost augmented with graph‑derived features against Graph Attention Networks, highlighting the value of explicit graph features for edge‑centric, imbalanced data. The results show that CatBoost with graph features achieves robust, interpretable performance, while GATs face overfitting in the imbalance; a regression task also demonstrates the model's capability to predict delay magnitudes. The authors also present a web‑based tool, Airdelay, for real‑time delay simulation, illustrating the practical potential to reduce delays and improve fuel efficiency and passenger experience.

Abstract

Flight delays due to holding maneuvers are a critical and costly phenomenon in aviation, driven by the need to manage air traffic congestion and ensure safety. Holding maneuvers occur when aircraft are instructed to circle in designated airspace, often due to factors such as airport congestion, adverse weather, or air traffic control restrictions. This study models the prediction of flight delays due to holding maneuvers as a graph problem, leveraging advanced Graph Machine Learning (Graph ML) techniques to capture complex interdependencies in air traffic networks. Holding maneuvers, while crucial for safety, cause increased fuel usage, emissions, and passenger dissatisfaction, making accurate prediction essential for operational efficiency. Traditional machine learning models, typically using tabular data, often overlook spatial-temporal relations within air traffic data. To address this, we model the problem of predicting holding as edge feature prediction in a directed (multi)graph where we apply both CatBoost, enriched with graph features capturing network centrality and connectivity, and Graph Attention Networks (GATs), which excel in relational data contexts. Our results indicate that CatBoost outperforms GAT in this imbalanced dataset, effectively predicting holding events and offering interpretability through graph-based feature importance. Additionally, we discuss the model's potential operational impact through a web-based tool that allows users to simulate real-time delay predictions. This research underscores the viability of graph-based approaches for predictive analysis in aviation, with implications for enhancing fuel efficiency, reducing delays, and improving passenger experience.

Figures (7)

• Figure 1: Example of airports and directed flights.
• Figure 2: Transformation of a multigraph of flights into a weighted directed graph. The multigraph (left) represents multiple flights between airports. In the weighted graph (right), edges are aggregated to show total flights as weights.
• Figure 3: Airport multigraph GAT Layer with multi-head attention for three different flights between nodes (SP,RJ), with alternating colors opacity for each flight.
• Figure 4: Feature importance for the CatBoost model on the airport network dataset, highlighting the relevance of graph-based features.
• Figure 5: Distribution of predicted delay values ($y_\text{pred}$) for the regression task using CatBoost.