Flight Delay Prediction at Hartsfield-Jackson Atlanta Airport: Insights from Machine Learning Models

Accurate prediction of flight delays is crucial for stakeholders in the aviation industry, as delays cost airlines billions annually in operational inefficiencies and passenger compensation, while disrupting connecting flights and airport resource allocation. This study employs advanced machine learning techniques to analyze arrival and departure delay patterns at Hartsfield-Jackson Atlanta International Airport (ATL), Georgia, USA, using ۲۰۱۹ flight data from the Bureau of Transportation Statistics. The analysis reveals that delays peak in June, while Saturdays experience fewer flights and shorter delays. Early morning hours (۵ AM–۸ AM) are identified as the most delay-prone period due to high traffic volume. Although Delta Airlines accounts for over ۶۰% of operations at ATL, Frontier Airlines recorded the highest percentage of delayed flights. Three classifiers—Logistic Regression, k-Nearest Neighbors (kNN), and Extreme Gradient Boosting (XGBoost)—were used to predict delays based on temporal and spatial factors. XGBoost achieved the highest performance, with an overall accuracy of ۸۳.۴۵% and a balanced accuracy of ۶۰.۶۷%. These findings highlight the importance of robust algorithms like XGBoost for high-traffic airports, where accurate delay prediction is critical for optimizing operations and enhancing passenger experiences. Temporal factors—such as the day of the month, month, and departure/arrival times—were the most significant predictors of delays, reflecting recurring patterns like peak travel periods and congestion. In contrast, location-based factors, including route distance and airport geography, had minimal influence, suggesting that delays are primarily driven by time-related operational dynamics rather than physical route characteristics. These findings underscore the value of machine learning in optimizing aviation logistics, enabling proactive delay management through data-driven insights.

کلیدواژه ها:

Delay Patterns ، Classification ، Flight Delay ، Atlanta Airport ، XGBoost ، Imbalanced Data