Heterogeneity mechanism and more precise prediction of seafarer fatigue based on group modelling

Traditional research often treats seafarers as a homogeneous group when exploring the causes of fatigue, neglecting the potential heterogeneity in fatigue mechanisms arising from differences in rank and department. This study systematically analyzes the fatigue driving patterns of four subgroups – Deck Department vs. Engine Department, and Officers vs. Ratings – and constructs high-precision prediction models. Based on questionnaire survey data from 450 seafarers from two Chinese shipping companies, Exploratory Factor Analysis was conducted to extract group-specific factors. Subsequently, Multiple Linear Regression and Backpropagation Neural Network models were established to identify key influencing factors and compare predictive performance. The results show that fatigue among Deck Officers is primarily driven by “Workload and Pressure" (β = 0.398); Engine Department seafarers (regardless of rank) are jointly influenced by “Sleep Quality" (β = 0.489-0.529) and “Work Pressure and Organizational Justice" (β = 0.415-0.451); Deck Ratings are most sensitive to “Sleep Quality and Environmental Interference" (β = 0.533). The predictive accuracy of the Backpropagation (BP) Neural Network model (test set (Formula presented)  = 0.636-0.895) was significantly better than that of the traditional linear model across all groups. The research demonstrates that seafarer fatigue exhibits significant group specificity, challenging the limitations of previous holistic studies, and provides a theoretical basis and effective tools for shipping companies to implement differentiated fatigue risk management.