Using a Reverse Engineering Type Paradigm in Clustering. An Evolutionary Programming Based Approach

The aim of this work is to propose a novel view on the well-known clustering approach that is here dealt with from a different perspective. We consider a kind of a reverse engineering related approach, which basically consists in discovering the broadly meant values of the parameters of the clustering algorithm, including the choice of the algorithm itself, or even – more generally – its class, and some other parameters, that have possibly led to a given partition of data, known a priori. We discuss the motivation and possible interpretations related to such a novel reversed process. In fact the main motivation is gaining insight into the structure of the given data set or even a family of data sets. The use of the evolutionary strategies is proposed to computationally implement such a reverse analysis. The idea and feasibility of the proposed computational approach is illustrated on two benchmark type data sets. The preliminary results obtained are promising in terms of a balance between analytic and computational effectiveness and efficiency, quality of results obtained and their comprehensiveness and intuitive appeal, a high application potential, as well as possibilities for further extensions.