Abstarct: One of the first steps in the regional flood frequency analysis is assigning of gauging stations to regions that their stations are expected to have similar in flood producing mechanisms. This process is known as regionalization. Regionalization methods baxsed on cluster analysis can be useful in identifying groups of stations that their flood producing mechanisms are similar. In this study, performances of various procedures of cluster analysis such as hybrid clustering, fuzzy clustering using fuzzy c-means algorithm and clustering using self-organizing feature maps for regionalization of Sefidrood watershed in order to perform regional flood frequency analysis have been assessed. Also assessment of homogeneity of regions and regional flood frequency analysis has been performed by using L-moments. Results showed that combination of Ward algorithm and K-means algorithm would be best option among hybrid clustering algorithms. Also results showed that fuzzy c-means algorithm has an acceptable performance in forming homogenous regions and obtaining appropriate estimates in regional flood frequency analysis using L-moments baxsed algorithm in interested watershed. In addition, this point has been observed that using of fuzzy clustering can provide reliable flood estimates for longer return periods. In the case of self-organizing feature maps this has been observed regions obtained by this technique in 3 regions state, provide suitable values for cluster validity measures and heterogeneity measures.