Abstract: This study develops and compares machine learning models to predict flight delays at Tirana International Airport (LATI). Using Neural Networks (NN), XGBoost, and Decision Trees (DT) as a baseline model, the study classifies flight delays in real time from the moment of departure. The models predict arrival delays, utilizing operational, environmental, and historical flight data. The performance of the models was evaluated at various thresholds, with XGBoost showing the best balance of precision and recall, while Neural Networks excelled in capturing delayed flights at the cost of increased false positives. The results show the use of machine learning algorithms in the management of delays for airports and airlines. Future work could improve the models through incorporation of additional airport features that can be more specific, like the number of runways, air traffic control information, and weather components like thunderstorms, strong winds, and high-altitude weather along the path of the flight, many of which might improve delay prediction and operational decision-making and might be captured easily through standard weather data.

Keywords: Flight Delay Prediction, Machine Learning, Neural Networks, XGBoost, Decision Trees, Real-Time Classification, Airport Operations, Tirana International Airport, Predictive Modeling, Delay Management

Pages: 580-590