Classification of “kelurahan” and rural area into urban/rural class basically meant to form a layer (stratum) were used in the survey sampling techniques. With the status of urban and rural areas, the sample can represent the entire population correctly. Proper selection of variables could distinguish village into urban and rural class. The purpose of this study was to provide an alternative selection of the most influential variables to determine the classification of villages in Indonesia with a mix method of bootstrap and binary logistic regression. The data used in this case is data Potensi Desa (PODES) 2011 which conducted by Badan Pusat Statistik. The methods used in this study are binary logistic regression and bootstrap. Logistic regression is one method of non-parametric regression where the response variable is categorical data. This method can also be used for data classification. Bootstrap, is known as one of the data simulation method, intended to simplify the inferential statistical analysis but produces a more robust analysis. From previous studies showed that the variable density of population, the number of farm households, and the presence of the primary facility is the most influential variables in the classification of villages in Indonesia. From the previous studies also can be concluded that the bootstrap approach give small mistake of goodness in variance covariance matrix. The more bootstrap replication is used, the more robust the resulting analysis. The results showed that the presence of variable existence of Junior High School and hotels can be removed from the model without effecting goodness of fit of the model. The addition of new variables, existence of the internet cafe and bank is able to produce more powerful model for classification of the village.