The zero-inflated binomial and binomial hurdle models are two models to address the problem of excess zeros for proportional data. Both model structures are similar, which are composed of two distributions. A degenerate process is responsible for producing zeros, which are called structural zeros. The other distribution could be a binomial or zero truncated binomial distribution which depends on the type of model in this study. Zero-inflated binomial (ZIB) model involves the sampling zeros. On the other hand, binomial hurdle(BH) model address the structure zeros only in the zero-inflated part, which is a difference between the two models. Generalized fiducial inference is a popular statistics analysis which is independent of frequentist school and Bayesian school. This method was first proposed by R.A. Fisher to challenge Bayesian school. The core idea is to switch the role of parameters and data, and to construct a distribution for parameters that contain all of the information of data. One significant advantage of generalized fiducial inference is that it does not rely on a prior distribution. In this research, the generalized fiducial inference is introduced to construct the confidence intervals for the means of zero-inflated binomial and binomial hurdle models and the formulation of confidence intervals are derived. A simulation will be conducted to test the performance of generalized fiducial inference on single parameters of ZIB model and BH model. Then the performance of generalized fiducial inference on the means of ZIB model and BH model will be tested through observing the coverage probability and mean width for the generalized confidence interval (GCI). We make a comparison between generalized confidence intervals and bootstrap confidence intervals. Two special cases of ZIB model and BH model are also considered. Generalized fiducial inference is also illustrated using the whitefly data. Through this data, we can test the performance of generalized fiducial inference under different numbers of trials and sample sizes. We construct generalized confidence intervals and bootstrap confidence intervals to show an excellent property of the generalized fiducial inference.