Related Experiment Video
Updated: May 27, 2026
05:37
An R-Based Landscape Validation of a Competing Risk Model
Published on: September 16, 2022
Machine learning model for multi-factor risk prediction in airport construction under non-stop operations
1Guangxi Airport Management Group Co., Ltd, Nanning, 530000, China. culm7602@outlook.com.
Scientific Reports
|April 4, 2026
Summary
Airport construction safety is improved using a data-driven approach. XGBoost and SHAP analysis identified key risks like flight density and visibility, enabling better risk management systems.
Area of Science:
- Aviation Safety
- Construction Management
- Data Science
Background:
- Airport construction during ongoing operations faces complex safety challenges.
- Traditional safety methods struggle with multi-factor interactions in dynamic airport environments.
Purpose of the Study:
- To develop a quantitative risk assessment model for airport construction safety.
- To identify critical risk factors and their impact on safety incidents.
Main Methods:
- A risk factor system with 42 variables across six categories was developed.
- Synthetic Minority Over-sampling Technique (SMOTE) addressed class imbalance.
- XGBoost classifier and SHapley Additive exPlanations (SHAP) were used for modeling and interpretation.
Main Results:
- The XGBoost model achieved 92.7% accuracy, 0.875 AUC, and 85.7% recall.
- Identified key risk factors: flight density, visibility, NOTAM release timeliness, peak hours, and personnel experience.
- Determined critical thresholds for flight density (35/hr), visibility (3 km), and NOTAM delay (2 hr).
Conclusions:
- The study provides a data-driven foundation for operational risk warning systems in airport construction.
- Findings support differentiated risk management strategies to balance efficiency and safety.
- SHAP analysis revealed synergistic effects of operational pressures on safety outcomes.
