Abstract
In this paper, we propose a method for decision of class borders. The method decides approximately the border on SOM from nearest neighbor data set of different classes. The method also decides the border on the SOM using the characteristic of SOM which can map a multi-dimensional data set to the low dimension which is 2 or 3 dimensions. Using two examples, it can be shown that the proposed method can improve the accuracy of Cluster analysis and show that the method makes the class border visualized accurately and comprehensible as well. One example is two-components data set where the class border can be easily plotted. The other is the iris data set where the class border is really difficult to be expressed.
