From Evolution of Organisms Cognitive Structures to Data Clustering. A Bio-inspired Approach to Build-up an Efficient Clustering Algorithm

Data clustering consists in finding homogeneous groups in a dataset. According to their own experience and background, many eminent Authors have proposed different definitions for data clustering. Here we propose a very general one that can cover almost all the data clustering approaches. Given a set of n points in a p dimensional space, cluster analysis aims at grouping data into k groups that are maximally homogeneous within each one and maximally heterogeneous between them (to generalise to any k the word "among" would be more appropriate, however, for historical reasons, the word "between" is more widely used to refer to the heterogeneity among groups, even if k>2). The importance attributed to cluster analysis is related to its fundamental role in many knowledge fields. Data clustering techniques are the “host ghosts” of many innovative applications for a wide range of problems (i.e.: biology, marketing, customer segmentation, “intelligent” machines, etc.). On the other hand, in wide epistemological terms, human and animal learning/adapting processes can be viewed as clustering methods to build-up concepts, mental categories, perceptual patterns, behavioural strategies and so on.