This research aims to group regencies/cities based on education indicators and identify the characteristics of each group formed based on education indicators. The method used in this research is Self self-organizing map (SOM). SOM is an artificial neural network that requires no assumptions and a method that produces a representation of the input space from low-dimensional training samples. The data used in this research are 9 variables regarding pure enrollment rates, gross enrollment rates, and student-to-teacher ratios at each level of education in 24 districts/cities in South Sulawesi in 2020-2021 which come from BPS publications. Based on the results obtained, 4 clusters were formed, each of which had its characteristics. The clusters formed include Cluster 1 consisting of 7 regencies/cities, cluster 2 consisting of 10 regencies/cities, cluster 3 consisting of 4 regencies/cities, and Cluster 4 consisting of 2 regencies. Based on the results of cluster validation using the Dunn index, 4 optimal clusters were obtained with a value of 0.42.