Prediction of air traffic delays: An agent-based model introducing refined parameter estimation methods

Area of Science:

• Air traffic management
• Complex systems modeling
• Predictive analytics

Background:

• Accurate prediction of flight delays is crucial for efficient air traffic operations.
• Existing models often lack detailed parameter estimation for individual flight behavior.
• Agent-based modeling offers a promising approach for simulating complex network dynamics.

Purpose of the Study:

• To develop and validate an agent-based model for predicting individual flight delays.
• To incorporate detailed parameter estimation methods for enhanced prediction accuracy.
• To analyze the impact of prediction horizon and model parameters on delay forecasting.

Main Methods:

• Developed an agent-based model with detailed parameter estimation for flight delay prediction.
• Utilized a conditional probability model for expected departure time adjustments.
• Employed random forest regression for estimating turnaround and elapsed flight times.
• Trained models on 2017 US flight data and tested on 30 days of 2018 data.

Main Results:

• Achieved an average absolute error of 6.8 minutes in delay prediction.
• Attained 89.5% classification accuracy with a 15-minute threshold for a two-hour forecast.
• Demonstrated superior performance compared to existing air traffic delay prediction research.
• Quantified the positive impact of parameter models and the negative impact of prediction horizon.

Conclusions:

• The proposed agent-based model with enhanced parameter estimation significantly improves flight delay prediction accuracy.
• The model provides a robust tool for air traffic management, offering valuable insights into delay factors.
• Further research can explore optimizing prediction horizons and incorporating real-time data for even greater precision.