Hierarchical Method of Self Clustering Algorithm Using Self Organizing Map Neural Network

This paper describes an approach of self clustering algorithm using hierarchical Self Organizing Map neural network. In the basic SOM algorithm the user should enter the number of clusters before its training, here we give different approach of representing only the data set and the network should create clusters according to the data topology. In order to do it there are several steps that have to be done. At first the data is clustered according hierarchical approach in this way we obtain data set tree, where each node represents a cluster and each cluster includes some data from the data set. The method is hierarchical[3], [4] because it starts from one cluster if one cluster is not enough then clusters are increased into two and so on until the algorithm stops creating clusters. The term self clustering means that the number of clusters are not determined before learning trial their number depends on the data and the algorithm determines their number.