Abstarct: One of the methods for estimating floods quantiles in ungagged catchments is the use of regional frequency analysis method. In conventional regionalization methods, hydrologic homogeneous regions are created baxsed on the characteristics of the catchment area, such as physiographic characteristics of the area, geographic location and flood statistics at each station. The identification of hydrological homologous regions is one of the fundamental issues of hydrology in two dimensions: applied and research. Using clustering methods is one of the common methods for achieving hydrological homogeneous regions for regional flood frequency analysis. In this research, the ability of self-organizing feature map algorithm to regionalization of the Karkheh catchment is investigated. The homogeneity of each obtained regions was evaluated by the Hosking and Wallis homogeneity statistic. Homogeneity statistics for the first and second clusters were calculated to be 0.33 and 0.17, respectively, indicating the homogeneity of each region. Also, ZDIST statistics for stations in each of the two hydrological homogeneous regions were calculated using the R programming language for 3 parameters distributions, and the Pearson type III and generalized extreme value distribution have been selected as the best regional distribution. The parameters of this distribution were calculated by linear moment method, and then flood with different return periods at each station was estimated using regional distribution. According to the results of this research, the clustering ability of self-organizing feature map can be used to reach the hydrological homogeneous regions elsewhere in the country.