Tracing delay network in air transportation combining causal propagation and complex network

In air transportation, monitoring delays and making informed decisions at a system level is crucial for network managers. Causal selection methods have recently witnessed increased adoption for the analysis of multi-observations. Systematic Path Isolation (SPI) stands out as an effective mechanism for selecting causal pathways in time-series data. However, specific improvements are needed to ensure the effectiveness within the aviation system. This paper proposes an SPI-based causal inference method that incorporates the Granger test and the Kernel-based test, accommodating both linear and non-linear relationships, thereby enabling better condition selection. Additionally, the two-step SPI test employs the Kernel-based Conditional Independence test due to its suitability for handling complex data with nonlinear relationships, and it avoids explicit feature extraction. Validation of delay tracing involves the use of complex network metrics and a specially designed load-embedded metric for identifying daily states. The case study results demonstrate the effectiveness of the network generated by the proposed method in accurately tracing dynamic states, particularly through the proposed indicator. In static propagation detection, network motifs can serve as micro-expressions, particularly with convergence and transmission forms during high delays. This research contributes to refine the depiction of delay propagation in the air transport network, enhancing the ability to trace delay trends in dynamic and static perspectives.