The present article dedicates itself to fuzzy modelling of data-inherent structures. In particular two main points are dealt with: the introduction of a fuzzy modelling framework and the elaboration of an automated, data-driven design strategy to model complex data-inherent structures within this framework. The innovation concerning the modelling framework lies in the fact that it is consistently built around a single, generic type of parametrical and convex membership function. In the first part of the article this essential building block will be defined and its assets and shortcomings will be discussed. The novelty regarding the automated, data-driven design strategy consist in the conservation of the modelling framework when modelling complex (nonconvex) data-inherent structures. Instead of applying current clustering methods the design strategy uses the inverse of the data structure in order to created a fuzzy model solely based on convex membership functions. Throughout the article the whole model design process is illustrated, section by section, with the help of an academic example.

This article proposes knowledge-based short-time prediction methods for multivariate streaming time series, relying on the early recognition of local patterns. A parametric, well-interpretable model for such patterns is presented, along with an online, classification-based recognition procedure. Subsequently, two options are discussed to predict time series employing the fuzzified pattern knowledge, accompanied by an example. Special emphasis is placed on comprehensible models and methods, as well as an easy interface to data mining algorithms.

Zugl.: Zürich, Univ., Diss., 2008

Zugl.: Karlsruhe, Univ., Diss., 2005

Zugl.: Cottbus, Techn. Univ., Diss., 2005

Zugl.: Duisburg, Essen, Univ., Diss., 2008

Zugl.: Hohenheim, Univ., Diss., 2007

Zugl.: München, Universiẗat der Bundeswehr, Diss., 2008.

Stuttgart, Univ., Studienarb., 2004.

Zugl.: Köln, Univ., Diss., 2007