Objective To screen candidate preparations of live rotavirus(RV) vaccine using stability data modeling.Methods The citric acid, sodium citrate and disodium hydrogen phosphate buffer bulk solutions were prepared, and sucrose and RV bulk solution were added after sterile filtration to prepare three RV liquid formulations: FA(5 × sucrose-1 × citric acid-1 × sodium citrate), FB(5 × sucrose-1 × citric acid-1 × disodium hydrogen phosphate) or FC(1 × sucrose), with MEM as the diluent and virus titer of 7. 5 logCCID_(50)/mL. The stability data of the three formulations were linearly extrapolated using 2-8 ℃ data extrapolation model and Arrhenius model, and compared with the actual stability curves at 2-8 ℃ for36 months to evaluate the predictive models and identify the optimal formulation.Results The two model analysis indicated that at temperatures ranging from 2 to 8 ℃, formulations FA and FB exhibited slower declines in virus titer, demonstrating significantly superior stability compared to FC. Both models confirmed the stability hierarchy as FA = FB > FC, with predictions based on medium-to-long-term data showing good agreement with 36-month real-time stability results.Conclusion This study successfully identified the more stable RV vaccine candidates FA and FB by establishing and validating two stability predictive models. These two models provide reliable tools for assessing the stability and efficiently screening RV vaccines.