No abstract

A constructive interval model of making a collective decision by an independent group of experts is developed. The model is based on a priori information about the frequency of experts' errors in estimating a random state of an object using a finite sample.Methods of collective decision-making are widely used in various fields [1][2][3]. Different approaches to constructing models of collective decisions are known, in particular, voting [4], ranking [5], averaging partial decisions of experts [6], algorithms of fuzzy rules [7], etc.Bayesian models of collective decision-making by a group of independent experts under conflict conditions are proposed in [8]. However, these models employ a priori knowledge of the probabilities of experts' errors, which are often unknown in practice.The development of interval analysis methods [9, 10] makes it possible to pass from point models to interval models based on Bayesian inference mechanisms [11,12]. The purpose of the present paper is to develop (in the context of interval analysis) Bayesian models of collective decision-making in conflict situations, based on a priori knowledge of the frequency of errors made by experts using finite experimental samples. PROBLEM STATEMENTLet a plant be in one of two possible states: V 1 or V 2 , randomly passing from one state to the other with a priori probabilities P V P V P V ( )and ( ) ( ) 1 2 1 1 = -. Two experts, A 1 and A 2 , using independent data, make a decision d i on the current state of the plant in the form of indicator functions:It is clear that the set of possible situations consist of four combinations of partial decisions (1). These decisions are agreed only in two cases (when the experts make the same decision), and in the other two cases the decisions are contradictory:S 12 1 2Let experts' error rates be estimated using a representative sample of n observations with known states of the plant:where E A i is the number of cases where the ith expert has made an incorrect decision.

No abstract

No abstract

No abstract