Because of the river network generalization is limited to such influence of factors as mapping usage, mapping scales, mapping areas and mapping data, how to select an appropriate number of rivers has been concerned by cartographers. Therefore, to decide the selective index of river network on the map has presently become a crucialissue of map compilation. The research plans to sample forty major and minor rivers in Taiwan to measure theriver order. The model of river generalization regression is induced with a statisticalmethod. By studying and analyzing a great number of the samples, the type of river network generalization in Taiwan can be hopefully induced and is expected to be a new thinking direction for the generalization of river network.