Multi-Class Flight Delay Prediction Based on an Improved XGBoost Model

Flight delay prediction is a critical issue in aviation management. This paper proposes a multi-class flight delay prediction model based on an improved XGBoost, aimed at achieving precise classification of flight delays. The study utilizes 2020 flight operation data from a certain U.S. airline, analyzing and processing the raw data to produce a standardized dataset. In light of the class imbalance within the dataset, a novel loss function is designed for optimization, and feature engineering is employed to enhance the model’s expressive capabilities. Furthermore, to efficiently find the optimal parameter configuration in XGBoost—given its numerous adjustable parameters—Bayesian optimization is introduced. Comparative analyses with various models validate the effectiveness and superiority of the proposed model.