A soft (fuzzy) clustering algorithm is used to divide the precipitation region of China into n clusters according to the precipitation amount, based on the data from 160 meteorological stations of China in the last 50 years, with each station having a degree of belonging to clusters, as in the fuzzy logic, rather than belonging completely to just one cluster. The effect of clustering parameters on regionalization is examined. The further study of the conventional 5category precipitation regionalization shows that some regions allocated into a specific category do not exactly belong to one certain cluster, but has some similarity with the other regions, such as Tongliao and Dalian, etc. The soft (fuzzy) clustering algorithm can provide more information on the study of the factors affecting the precipitation to better describe the similarity, and be able to obtain the agreement with the results from other classification approaches after degradation.