Background Hepatocellular carcinoma (HCC) rupture is a life-threatening complication associated with poor prognosis. This study comprehensively analysed risk factors for HCC rupture and developed a predictive model supplemented by machine learning models for early risk identification and clinical decision-making.Methods This retrospective study analysed patients with and without HCC rupture from tertiary centres in China between January 2016 and June 2019. Propensity score matching (PSM) was used to reduce baseline differences between the rupture and non-rupture groups. Random forest and deep learning models were developed to enhance predictive accuracy and interpret variable importance. Model performance was evaluated using metrics such as precision, recall, and the F1 score across training, validation, and test cohorts.Results Among the 5952 HCC patients, the median follow-up duration was 48.6 months. Key risk factors for HCC rupture identified in this study include cirrhosis, protrusion ratio, and tumour maximum length. The CAPTure nomogram, constructed based on these predictors, yielded area under the curve (AUC) values of 0.857, 0.824, and 0.840 in the training, validation, and test cohorts, respectively. In the test cohort, the random forest and deep learning models achieved AUCs of 0.870 and 0.872, respectively.Conclusion This study provides a comprehensive analysis of risk factors for HCC rupture and introduces the CAPTure model as a practical and accurate tool for clinical use. By integrating traditional and machine learning approaches, the findings of this study offer robust methods for early risk assessment, resource optimization, and improved management of HCC rupture. This study comprehensively analysed risk factors for hepatocellular carcinoma rupture in a nationwide Chinese cohort and developed the CAPTure (Cirrhosis, Assessment of Protrusion ratio, and Tumour maximum length) predictive model. The model demonstrated strong predictive accuracy (area under the curve 0.857-0.840) across training, validation, and test cohorts. The inclusion of machine learning techniques further enhanced prediction capabilities, emphasizing early risk identification and personalized interventions to improve clinical outcomes.