DEVELOPMENT OF FLIGHT DELAY PREDICTION SYSTEM USING SUPERVISED MACHINE LEARNING

• Main Author: Saiful Anwar, Muhammad
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/82297
• Institution: Institut Teknologi Bandung

Flight delays are a significant and escalating issue as the aviation sector records continuous growth. These delays cause losses for passengers, airlines, and airports. Although several flight delay prediction models have been developed, their performances remain rather low. This research aims to create an explainable artificial intelligence (XAI) flight delay prediction model using machine learning. The methods employed in the research include formulating delay categories, comparing several machine learning methods, selecting the best model, and conducting feature influence analysis using the SHAP (Shapley Additive Explanations) method. The data employed is domestic flight data in the United States from 2021 to 2022. The research results in a prediction model using the random forest algorithm, achieving an accuracy of 83% and an F1 score of 0.83. SHAP analysis identifies key factors influencing delays, including turnaround time, previous delay duration, airline, airport delay, and flight distance. These findings are beneficial for airlines to improve on-time performance (OTP) and for airports to enhance slot management.