A hierarchical clustering method for attribute discretization in rough set theory

In this paper, hierarchical clustering is introduced. The method can determine automatically the significant clusters in a hierarchical cluster representation. It could choose best classes for discretization by scatter plots of several statistics primarily. Moreover we can extract the clusters from dendrograms that contain essentially the same information, which shows the two discretization results are consistent. By comparison among several cluster algorithms with the defect inspection of wood veneer, hierarchical clustering discretization method is typically more effective and advisable.