Abstract: The cluster analysis method is used in this paper to classify 30 releves (292 species) of evergreen broad-leaved forest which collected from different places in Guangxi. The distance formula Euclidean is adopted for calculation of similarity between releves. Eight’ cluster methods: nearest neighbour, furthest neighbour, median, centroid, group average, flexible group average, incremental sum of squares and flexible cluster method, are used in cluster analysis.The results of using cluster analysis are similar to, and as well as different with the classification by dominance-type. However, the cluster analysis may differentiate the releves in more detail. In some cases it can not only incorporate the releves, which come from different places and have same dominants, into same type, but also divide them into distinct groups. Sometimes it can also put the coming from same place and having different dominants releves together. All of them reflect the floristic composition which relate to latitude and altitude and other habitat factor to be changed. Based on the results of cluster analysis, 25 releves are divided into seven types and the diagnostic species are sought out of the five types of them. In addition, eight cluster analysis methods and threshold value of classifying levels are discussed.