| Abstract: |
Flight delays remain one of the most persistent and economically debilitating challenges confronting the global aviation industry, and the Kingdom of Saudi Arabia's rapidly expanding aviation sector driven by Vision 2030 targets, the National Aviation Strategy, and surging Hajj and Umrah passenger volumes is no exception. Despite substantial infrastructure investment, including the expansion of King Khalid International Airport (Riyadh), King Abdulaziz International Airport (Jeddah), and King Fahd International Airport (Dammam), airlines continue to suffer from cascading delay propagation that disrupts operations, erodes passenger satisfaction, and incurs significant annual losses. This empirical study investigates the application of predictive analytics to mitigate and proactively manage flight delays across Saudi Arabian airline operations. Using a dataset comprising 1.9 million flight records collected from Saudi and international carriers operating in Saudi airspace over a five-year period (2018–2023), the study applies multiple machine learning algorithms Random Forest, XGBoost, Long Short-Term Memory (LSTM) networks, and Gradient Boosted Trees to predict delay probability and duration with high accuracy. Empirical results demonstrate that the XGBoost model achieves a prediction accuracy of 92.1%, outperforming baseline logistic regression models by 24.1 percentage points. The study identifies sandstorms and reduced visibility, extreme summer heat-related operational constraints, air traffic congestion during Hajj/Umrah peak periods, and aircraft turnaround time as the dominant causal factors of delay. Additionally, the integration of real-time ADS-B data streams with historical operational data is shown to reduce mean absolute error in delay prediction by 36%. |
