The Concept of Reverse Clustering

The paradigm of the reverse clustering is presented, consisting in attepting to reconstruct, via some clustering procedure, of a definite partition of a given set of data. We give some examples of such real-life problems. The reverse clustering problem is formulated, and its component parts described, i.e. the clustering algorithms considered, their parameters, the weighing or selection of variables, characterising the data items, the definition of distance, and, finally, the measure of similarity of the given definite partition and the one obtained from the reverse clustering procedure. Notation used throughout the book is also introduced. Some comments are given on the applied search procedure.