Machine Learning-Based Prediction of Passenger Waiting Time During Check-in Process

Abstract

<p>The check-in process is a crucial aspect of airport management, requiring effective coordination between the terminal and airlines. Emergencies and the pandemic have exacerbated challenges in managing the check-in process, resulting in long queues and extended waiting times, particularly during peak departure periods. Predicting check-in waiting times accurately can optimize terminal operations and enhance passengers’ departure experience. Therefore, there is an urgent need for airports to possess predictive capabilities to fully leverage their facilities. This paper presents a machine learning-based approach for predicting passenger check-in waiting time. Firstly, this paper collects real data from one of the largest worldwide airports in its major domestic terminal from September 2021 to January 2022. Next, the collected data is analyzed and processed, with continuous features categorized to derive meaningful response variables. Moreover, this paper compares various machine learning classifiers and optimizes the best-performing classifiers, such as Gradient Boosting Machine (GBM) and Random Forest (RF), and discusses the impact of thresholds and features on the accuracy of the models. Based on real-world data analysis, Gradient Boosting Machine exhibits the highest multi-class classification accuracy (0.790; 0.731) and F1-score (0.648; 0.479) compared to other models, achieving an overall AUC of 0.95. The experimental findings suggest practical applications for airport management in both current and future prediction scenarios. This model has been applied in the airport system to facilitate the rational allocation of check-in resources.<br></p>