Abstract: A Bayesian discriminant analysis method based on a continuous ratio Logistic regression model is proposed to address the limitations of the traditional Bayesian discriminant analysis method in dealing with practical problems. Firstly, a semiparametric density ratio model is established based on the continuous ratio Logistic regression model, and the parameters of the model are estimated by the empirical likelihood method, and the posterior probabilities are calculated using Bayes' theorem for classification prediction. Statistical simulations then compare the correct rate of judgement between the new method and the traditional method. Statistical simulations show that the discriminative power of the two is comparable when the overall data conforms to a normal distribution, otherwise, the proposed new method can better discriminate different data characteristics. Finally, the new method is applied to analysis real data sets to verify the accuracy and robustness of the new method in classification prediction, which is more suitable for modelling and prediction of multivariate classification problems in practical applications compared with the traditional method.