SearchPublisher and Institutes
You are here: Home :: Area NEM :: Engineering :: Automation
Peter Krause, T. Slawinski
The Fuzzy–ROSA Method: From the Rule–Oriented Statistical Analysis to Data–Based Generation of Interpretable Takagi–Sugeno Systems
The use of data mining methods, especially data–based modeling, in the field of automation has experienced a boom over the last years. On the one hand the users are in need of models for systems with growing complexity. On the other hand there are new possibilities of solving problems, where common solving strategies failed, because of the increasing capacity of computer systems and the availability of new methods, especially in the field of computational intelligence. In many cases interpretable models are required to be accepted in industry. Moreover, interpretable models have the advantage that available knowledge can be easily integrated and with data–based methods new information can be gained. A method which has proven to be very efficient when it comes to generate automatically compact and interpretable small fuzzy rule bases is the Fuzzy–ROSA method. In this paper we show the present state of the Fuzzy–ROSA method and then we present in detail current research activities, which led to a far wider area of applications. We demonstrate how new test and rating strategies consider different modeling objectives and how the efficiency of the rule search is increased by new search methods. At last we present in detail an extension to the Fuzzy–ROSA method which generates interpretable Takagi–Sugeno systems with high modeling accuracy.
at – Automatisierungstechnik, Oldenbourg Wissenschaftsverlag
Print ISSN: 0178-2312
Volume: 49, 09/2001
Pages: 391