Objective To investigate the data fitting model of ELISA for determination of gentamicin residues in attenuated Japanese encephalitis vaccine (JEV), so as to validate and optimize the quantitative detection system. Methods The JEVs were tested for gentamicin residue using commercial gentamicin residue ELISA kit, and the test data were fitted respectively through quadratic polynomial model, log-linear model and four-parameter logistic model. Then the indicators including the coefficient of determination (R2), experimental point distribution and recovery rate were analyzed and compared. Results The variables tests of quadratic polynomial model, log-linear model and four-parameter logistic model were all of statistical significance (P<0.05), and the mean values of R2 were 0.997 0, 0.994 4 and 0.999 3, respectively, among which the quadratic polynomial model and four-parameter logistic model presented the best fitting. The distribution of standard points on both sides of the standard curve was uniform for all models, and the four-parameter logistic model was the best, especially in the range of medium and low concentrations (0.1-0.9 ng/ml). The recoveries of standard substances in the concentration range of 0.1-8.1 ng/ml of each fitting model met the requirements of Chinese pharmacopoeia, and the differences were not statistically significant. Conclusions Quadratic polynomial model, log-linear model and four-parameter logistic model are all suitable for the data analysis of ELISA for gentamicin residues, but the quadratic polynomial model and four-parameter logistic model fit well, which has important guiding significance for continuous optimization of this quantitative determination.