A Novel Prediction Model of Passengers Arrival at Airport Check-in Counters

Abstract

This study proposes a novel model to predict the advance arrival of passengers at airport check-in counters. While individual flight arrival patterns vary significantly and are challenging to predict, their cumulative arrival patterns are comparatively more consistent. Leveraging this insight, the study predicts key time points of the cumulative arrival curves and uses the estimated curves to derive the arrival patterns. To evaluate the model's performance, arrival data from 33 flights operated by the study airline at Terminal 2 of Taoyuan International Airport in Taiwan were collected. Three key time points—the 5^th percentile (t₁), 50^th percentile (t₂), and 95^th percentile (t₃) of the cumulative arrival curves—were regressed against the number of passengers, flight length, season, and time of day. The proposed model achieved a corrected Mean Absolute Percentage Error (CMAPE) of every 10 minutes 38.4% and every 30 minutes 13.5%, outperforming the commonly used clustering approach (48.0% and 28.2%). These results indicate the proposed model's satisfactory predictive performance.