Employing immune network model for clustering with plastic structure

A clustering method that generates a plastic cluster structure is proposed by employing the immune network model. Various kinds of clustering and categorization methods have been applied to the information visualization systems on WWW. However, the user´s context through a series of information retrieval is not fully considered. The proposed clustering method can reflect the user´s context to the cluster structure by reusing the clusters that have been effective in the previous retrievals. The behavior of the proposed clustering method is analyzed with preliminary experiments, and it is shown that the set of clusters can be activated without overlapping. The function of the memory cell is also introduced, which enables one to give a priority of activation to a specified cluster.