Ein Konzept zur automatischen Lösung von Entscheidungsproblemen bei Unsicherheit mittels der Theorie der unscharfen Mengen und der Evidenztheorie

Abstract

Vorwort AbstractNumerous methods allow for the automatic design of decision systems with respect to uncertainty. Among these are machine learning, artificial neural networks and statistical methods. Only very few cases consider the different consequences of false decisions. Most design methods also fall short of the possibility to delegate decisions to an expert in cases of contradictive or insufficient information.This thesis contributes to the advancement of methods for the automatic design of fuzzy decision systems. These decision systems are assessed using decision theory. A cost matrix is used to model the consequences of false decisions. The attribute costs used by the decision system are considered optionally in the decision system's assessment. References derived from decision theory are developed for all design steps of fuzzy systems. Their use is demonstrated with a set of demonstrative examples. A new interference method to delegate decisions to an expert is developed using evidence theory. The design methodology for fuzzy systems is extended by a procedure to obtain multi-stage rule bases. Such rule bases do not evaluate all criteria for each decision and thus contribute to lower attribute costs.Benchmark data sets and a realistic robotics example are used to demonstrate the method's performance. The advantages of the interpretability of fuzzy systems is explained using an example.

Assessing natural hazards in forestry for risk management: a review

“…Therefore, a sound theoretical framework has to be built, using adequate theories like the Dempster-Shafer theory of evidence (Shafer 1976) or the theory of fuzzy sets (Zadeh 1965). A general overview of the theoretical background can be found in Beck (2005). Both theories have already been applied to problems of decision making in forest management.…”

Stichwort: Unsicherheit — Ungewissheit — Entscheidungen

“…z.B. ZIMMERMANN 1993;ROMMELFANGER 1994;BECK 2004). Die Bandbreite der Kategorisierungen reicht von ‚formal-normativen' über ‚probabilistische' bis hin zu ‚integrativ-ökologischen' Ansätzen.…”