<![CDATA[This study developed an MSI-transformed multiple linear regression model to estimate seafarers’ safety compliance from fatigue and occupational stress and implemented the resulting equation in a Python-based Maritime Intelligent Decision Support System prototype. Data were collected from 49 professional seafarers using a five-point Likert questionnaire measuring fatigue, occupational stress, and safety compliance. The ordinal responses were transformed into interval-based scores through the Method of Successive Intervals and analysed using ordinary least-squares regression. The overall model was statistically significant, F(2,46)=4.139, p=0.0222, with R2=0.153. Fatigue produced a negative but non-significant coefficient (β=−0.1020, p=0.489), while occupational stress also showed a negative coefficient (β=−0.2366, p=0.092). Thus, the predictors jointly accounted for a modest proportion of compliance variation, although neither showed an independently significant association at the 5% level. The estimated equation was embedded in the prototype to generate compliance scores and assign preliminary Safe, Alert, and Critical categories. Generative AI was restricted to translating deterministic outputs into concise mitigation narratives and did not calculate scores or determine categories. The study provides a transparent and reproducible workflow integrating ordinal-score transformation, interpretable regression, and computational decision support. The resulting architecture preserves traceability, supports human oversight, and demonstrates an exploratory early-warning framework for maritime safety compliance rather than a validated predictor of accidents or individual unsafe behaviour under current operational assessment conditions.]]>
