Airport time profile construction driven by flight delay prediction

Area of Science:

• Aviation Operations Research
• Data Science in Transportation
• Airport Slot Management

Background:

• Airport slot management balances market demand with operational capacity.
• Developing efficient 18-24 hour timetable profiles for busy airports is a significant challenge.
• Existing slot parameters must align with operational efficiency and market demand fluctuations.

Purpose of the Study:

• To develop a predictive model for airport time structure and flight delay levels.
• To integrate weather conditions and operational constraints into slot management.
• To enhance the reliability and efficiency of civil aviation slot scheduling.

Main Methods:

• Utilized historical flight and weather data for analysis.
• Applied K-means clustering and partial least squares regression for time structure modeling.
• Employed ensemble learning, specifically random forest, for flight delay prediction.

Main Results:

• Random forest demonstrated high accuracy in regression and prediction tasks.
• Successfully integrated time profile limits (based on weather) with delay predictions.
• Flights within the defined time parameter intervals averaged delays under 15 minutes.

Conclusions:

• The proposed model effectively predicts flight delays and optimizes timetable profiles.
• Achieving less than 15-minute average delays within predicted intervals enhances operational expectations.
• This approach offers a robust solution for managing busy airport slot schedules under varying conditions.