Semantic driven fuzzy clustering for human-centric information processing applications

Abstract:

This chapter presents an overview of fuzzy clustering techniques aiming at human-centric information processing applications and introduces the accuracy-interpretability tradeoff into the conceptualization of the clustering process. Nowadays it is a matter of common agreement that the cornerstone notion of information granulation is fundamental for a successful outcome of exploratory data analysis and modeling in fields like science, engineering, economics, medicine and many others. There is no doubt that fuzzy clustering is an excellent medium to obtain such information granules. For a matter of self-containment the chapter starts by presenting the fundamentals of fuzzy clustering along with some variants and extensions. In the second part of the chapter, the fuzzy clustering approach is highlighted as a valuable human-centric interface: the roadmap from data to information granules is displayed along with a discussion on some mechanisms to implement user relevance feedback. In the last part of the chapter a semantic driven evolutionary fuzzy clustering algorithm is analyzed, as a particular instance of a class of unsupervised clustering algorithms which embraces constraints usually applied in supervised learning. The results show that these more general constraints while tuning the equilibrium between accuracy and interpretability concomitantly help to unveil the structural information of the data. © 2009 Springer-Verlag Berlin Heidelberg.

Año de publicación:

2009

Keywords:

Fuente:

scopus