A top‐down hierarchical classification method
Hervé P. Hillion, Jean‐Marie Proth · Applied Stochastic Models and Data Analysis · 1987
Abstract In this paper we present an original top‐down hierarchical classification algorithm. In our approach we associate to each decomposition a ‘degree of separability’ which is used to evaluate the efficiency of the decomposition. In particular, it shows whether or not the classes are well separated and also whether or not they are homogeneous. This algorithm can be applied to items defined by real and bounded numbers.