<![CDATA[Right-censored data remain a persistent challenge in survival analysis, particularly in exclusive breastfeeding studies, where premature cessation is common, yet appropriate predictive models are rarely validated, risking biased estimates that misinform interventions. This study compares three survival analysis models, namely Weibull Accelerated Failure Time (AFT), Cox Proportional Hazards (CPH), and Random Survival Forest (RSF), in addressing right-censored data. Performance was evaluated through simulations varying sample sizes, censoring proportions, and hazard function shapes, using the Concordance Index (C-index) and Integrated Brier Score (IBS) with Inverse Probability of Censoring Weighting (IPCW). The simulation results showed that all three models achieved good discriminative performance, with median C-index values above 0.65 across scenarios. The Weibull AFT model performed best when the simulated data conformed to its underlying distribution. Both the CPH and Weibull AFT models consistently outperformed the RSF model, particularly under scenarios characterised by constant or decreasing hazard functions. Analysis of variance (ANOVA) confirmed that model type, sample size, and censoring proportion significantly influenced model performance. The Weibull AFT model was subsequently applied to real-world data on the cessation of exclusive breastfeeding in Donggala Regency. The findings indicated that psychosocial factors, including resilience, self-efficacy, and spousal support, served as protective factors that prolonged the duration of exclusive breastfeeding. In contrast, maternal fatigue and certain childbirth conditions were associated with an increased risk of early breastfeeding cessation. Sociodemographic variables, including age, educational attainment, and employment status, were not statistically significant predictors of breastfeeding cessation.]]>
