This paper fully dug the grammatical information of the rainfall as a variable, used the way of genetic algorithm clustering based on diversity sequence information entropy measure theory; and made regionalization research on the 99 precipitation sta -tions above the Bengbu Station in the Huaihe River Basin. The optimal classification was obtained based on the measures of therainfall sequences of the precipitation stations in the various sub-regions. Therefore, the research objective of high homogeneity a -mong the rainfall information in the sub-regions was achieved. Finally, the 3 classifications of 2, 4 and 8 were checked by usingthe accuracy of interpolation, and taking un-categorization as reference object. The results show that the classification can improvethe accuracy of rainfall interpolation.