Currently the government in poverty reduction programs are already starting to see the poor vulnerable population groups. The reason for the group are in a vulnerable position to be able to get into the group of poor people. In addition to the government's poverty eradication should not only globally but was based on the territorial aspect. The problem of poverty could have been influenced by the location and neighboring, so data observations difficult to be assumed independent. One of the spatial analysis are Geographically Weighted Regression (GWR). The weighting matrix for the GWR in this study was determined by adaptive bisquare function. The purpose of this study are to determine the characteristics of vulnerability households in Java and the variables that affect to it based spatial analysis with GWR. The result as follow that the influence of independent variabels in each districts/cities in Java vary significantly between regions. In addition the results of variation local coefficient produce that variabel KRT non migrants, KRT working in informal sector, and KRT low educated have significantly spatial heterogeneity in each districts/cities in Java. While large family variabels (the number of ART more than four people) and not own home is global.