Application of text mining and coupling theory to depth cognition of aviation safety risk

Aviation accidents have attracted a lot of attention due to the fatalities and considerable property loss. The complexity underlying these accidents is often the result of multiple risk factors coupling together, thereby increasing the difficulty of risks identification and comprehensive controlling. It is necessary to enhance proactive safety management capabilities and aircraft operational reliability through deeply cognitive risk information. This paper explores the application of Latent Dirichlet Allocation (LDA) model of text mining and coupling theory in two aviation safety datasets. The results of applying LDA topic modeling categorize risks across different datasets. Subsequently, the N-K model is employed to measure various coupling modes of risk factors, and the coupling effects between different risk factors were further analyzed. The results indicate that the two datasets identify different risk categories, and the coupling values of the risk factors usually increase with the number of factors. The results also reveal a higher degree of coupling among three risk factors: pilot manipulation, aircraft systems, and aircraft engines in the NTSB dataset. The proposed methodology shows significance in enhancing the accuracy of aviation risk analysis and the efficiency of accident prevention.