Diagnosis of Chongqing SF Air Express Delay Based on BN

Abstract

Aiming at the problems of complex causes for air express delivery delays of SF Airlines in Chongqing, the difficulty in accurate attribution using traditional experience, and challenges in accessing internal operational data of enterprises, this paper constructs a three-level node Bayesian network model of "Root Causes – Process Factors – Delay Outcomes" supported by theories of logistics timeliness management and Bayesian network fault diagnosis. The study adopts a hybrid data collection scheme combining public data extraction and business logic derivation. The conditional probability table is constructed through a three-step method: "qualitative grading → quantitative indicator mapping → probability derivation". Relying on the fullyear operational data of 2025 and 9 real delay cases, the model performance is verified from three perspectives: prediction, diagnosis, and sensitivity analysis.

The results show that the delay probability is the highest during the pre-delivery stock-up period of Double 11, and the insufficient systematic processing capacity caused by freight volume peaks constitutes the core contradiction of delays. The model achieves 100% diagnostic accuracy for single-link abnormalities in ground operation links, with an overall diagnostic accuracy of 66.7%. Sensitivity analysis indicates that freight volume peaks increase the delay probability by 7.1%, far higher than customs inspection, meteorological conditions, and air traffic control.

This study verifies the feasibility of a low-threshold modeling approach, providing a implementable tool for accurate delay diagnosis and timeliness optimization for air logistics enterprises, as well as a reproducible methodological framework for logistics timeliness research in data-scarce scenarios.