This paper considers the prior probability (PP) in classifying two populations by Bayesian approach. Specially, we establish the probability density functions for the ratio and the distance between two PPs that be supposed to have Beta distributions. Also, we build the posterior distribution for PPs when knowing the prior Beta distributions of them. From established probability density functions, we can calculate some typical parameters such as mean, variance and mode. According to these parameters, we can survey and also estimate the prior probabilities of two populations to apply for practical problems. The numerical example in one synthetic data set, four bench mark data sets and one real data set not only illustrate for the proposed theories but also present the applicability and feasibility of the researched problem.