“…Data clustering can be achieved through the topology representation of the high-dimensional data, that is, the data vectors with similar distribution characters are grouped together and displayed graphically thanks to the visualization advantage of SOM [4]. Thus, SOM naturally has many applications such as image and text segmentation [5], [13], medical diagnosis [6], [12], [14], target recognition [7], and traffic attack detection [8], [9], [20]. However, previous studies have theoretically predicted that the original simple model of SOM fails to fully represent the topology of the asymmetric-distributed (network traffic) data [4], [10].…”