The main requirement of regional frequency analysis is the homogeneity of the study area. The results of regional frequency analysis will be valid while the study area is homogeneous. Classifying an area into homogeneous regions causes more careful predicting of flood for different return periods at areas with no data or low-recorded length. Numerous methods are used for identifying of the Homogeneous Regions that the efficiency of these methods is dependent on identifying the effective basin variables and their weights. In this study, the canonical correlation analysis method using the pair canonical variables and its relations are able to determine the region of ungaged station. Based on the results, the regional homogeneity of ungaged station obtained by canonical correlation analysis method has more stations and the RMSE of flood estimation in ungauged sites is low in compare the Mahalanobis distance (MD) that it is caused increasing the length of the time record and the reliability in regional frequency analysis. As well as, the canonical correlation analysis method showed that there is an effective correlation between the watershed variables and flow variables and indicated the effectiveness of linear combination for area and p24 on flow variables. These variables can be used to model the flow variables in this area.