It is very difficult to judge the performance of a hierarchical cluster, but relatively
straightforward to measure the quality of a flat classification. The aim of a clusterer is
to produce low Sum Squared Error (SSE) but in a dynamic network this is complicated
by the ability of the net to use an unbounded number of nodes. A more reasonable
comparison for flat and hierarchical clusterers is, therefore, between the product of SSE
and number of nodes. The results are summarised in table 2, in which the column
marked “*” is the product of SSE and the number of nodes in each network.