Possibility theory and the risk. (Q647927)

Possibility theory (initiated by Zadeh) is an alternative to probability theory in the treatment of uncertainty. It studies those situations of uncertainty in which the events do not occur a large number of times and therefore the information is not extracted from a large volume of data. The development of possibility theory is due to a large number of authors, especially to Dubois and Prade. It has been successfully applied in decision making problems in conditions of uncertainty, in fuzzy cooperative games, fuzzy neural networks, etc. Possibility theory is based on new concepts such as possibility measure, necessity measure, possibility distributions, etc. Traditionally, probabilistic distributions are interpreted as fuzzy sets. Chapter 2 is a short introduction to fuzzy sets and fuzzy numbers. The central Chapter 3 is possibility theory, especially possibilistic indicators associated with fuzzy numbers. First, some notions and elementary concepts of probability theory are recalled. Then, fundamental concepts of possibility theory (possibility measure, necessity measure, possibility distribution, etc.) are defined and relations among them are studied. The substantial part of the chapter refers to the three possibilistic indicators associated with fuzzy numbers (expected value, variance, covariance). Chapter 4 presents the two possibilistic approaches to risk aversion. In a possibilistic context, notions corresponding to probability theory of risk aversion are defined etc. Two notions of possibilistic expected utilities are introduced and two distinct notions of possibilistic risk premium and possibilistic relative risk premium are derived from them as measures of risk aversion. Chapter 5 deals with expected utility operators. They are introduced axiomatically and the notions of possibilistic expected utility and possibilistic variance are defined. In this setting a general theory of possibilistic risk aversion is developed. Chapter 6 proposes a possibilistic model of risk aversion with several parameters. The notions of possibilistic risk premium vector and possibilistic risk premium associated with a possibilistic vector and a multidimensional utility function are defined. The two notions measure an agent's risk aversion to a situation of uncertainty with several risk parameters. Chapter 7 analyses a more complex risk situation: some risk parameters have a probabilistic representation and others a possibilistic one. The notion of mixed risk vector is obtained, in which some components are random variables and others are fuzzy numbers. For this case the notion of mixed expected utility is defined, from which the mixed risk premium associated with a mixed vector and a multidimensional utility function is defined. Chapter 8 deals with credibility theory initiated by Liu and Liu. Credibility theory is a way to model uncertainty phenomena. The credibility measure replaces the probability measure from probability theory and the possibility measure and necessity measure from possibility theory. In this chapter the properties of credibility measure, fuzzy variables and their credibilistic indicators (credibilistic expected value and credibilistic variance) are studied. In this context a credibilistic model of risk aversion is presented.