The data used in this study amounted to 15182 datasets consisting of 14 variables. Existing variables are divided into basic variables and additional variables. The basic variables consist of 5 variables namely Home ownership, Roof type, Wall type, Floor type, Defecation facilities. While the additional variables consist of 9 variables, namely employment, having money / livestock / jewelry deposits and others, welfare deciles, education, recipients of non-cash food assistance, recipients of productive assistance for micro enterprises, recipients of cash social assistance, recipients of family hope programs, and recipients of basic necessities. Using the naïve bayes algorithm classification method, the values of accuracy, precision, recall, and f-measure are 67.61%, 67.97%, 93.71% and 78.79%. The addition of additional variables to the basic variables resulted in an accuracy of 68.29% in the additional variables of education. This shows that by adding additional variables, the accuracy results are higher than using only basic variables, so that this study can be used as a recommendation in decision making on the implementation of determining the beneficiaries of the rehabilitation program for uninhabitable houses