Short-Term Forecasting of Checked Baggage Flow in Airports Using a Hybrid Improved Particle Swarm Optimization-Back Propagation Neural Network Model

Abstract

Accurate baggage flow (BF) prediction is crucial for airport and airline operations, enabling timely decision-making and efficient resource allocation. However, current approaches often estimate BF indirectly based on passenger flow (PF), failing to adequately capture the multi-timescale dynamic correlations between the two, which limits prediction performance. To address this gap, this paper proposes a hybrid modeling framework. First, the dynamic correlation between PF and BF is examined across four time granularities: annual, monthly, weekly, and daily. Quantitative analysis reveals a clear time scale dependency, clarifying the coupling complexity and key influencing factors of BF. Second, multi-dimensional data sources are integrated to scientifically select feature vectors for short-term BF modeling. Finally, a hybrid prediction model, named “improved particle swarm optimization-back propagation neural network” (IPSO-BPNN), is developed. This model incorporates an improved particle swarm optimization (PSO) algorithm to optimize a back propagation (BP) neural network for accurate short-term forecasting. A case study conducted at a major Chinese hub airport confirms the method’s effectiveness. Multi-metric evaluations demonstrate that IPSO-BPNN significantly outperforms existing methods, that is, BP, PSO-BP, and genetic algorithm (GA)-BP, when incorporating multiple factors, improving R² by 8.08%, 5.32%, and 5.40%; reducing mean absolute error by 25.92%, 21.56%, and 20.32%; and lowering root mean square error by 27.91%, 22.44%, and 19.14%, respectively. The findings provide practical decision support for resource management and optimization in baggage transportation systems.