Some quantities represented by the Choquet integral
Fuzzy Sets and Systems 56 (1993) 229-235 North-Holland
229
Some quantities represented by the Choquet integral Toshiaki Murofushi Department of Comm...
Fuzzy Sets and Systems 56 (1993) 229-235 North-Holland
229
Some quantities represented by the Choquet integral Toshiaki Murofushi Department of Communications and Systems, University of Electro-Communications, 1-5-1 Chofugaoka, Chofu, Tokyo 182, Japan
Michio Sugeno Department of Systems Science, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama 227, Japan Received April 1992 Revised August 1992
Abstract." This paper shows that the Choquet integral can represent some useful quantities such as supremum, infimum, essential supremum, essential infimum, mean, median, a-quantile (100a-percentile), and L-estimators including a-trimmed mean.
Keywords." Measure theory; probability theory and statistics; fuzzy measure; Choquet integral; Sugeno integral.
1. Introduction The purpose of this paper is to point out that the Choquet integral [2, 3, 6, 7, 12] can express several important quantities in probability theory and statistics, that is, the Choquet integral is very powerful to express statistical quantities. For instance, the Choquet integral can express the mean of a random variable; it is well-known that, for a given random variable X on a probability space (t2, ~, P), the Choquet integral of X with respect to P equals the Lebesgue integral of X with respect to P, i.e., the mean of X. The Choquet integral can express much more quantities. In the sequel we shall discuss this issue. Throughout the paper we assume that (t2, ~ ) is a measurable space. A fuzzy measure on (~, ~ ) is a real-valued set function ~ : ~ [0, 1] satisfying (1) It(0) = 0 and/z(I2) = 1, (2) /z(A) <~/x(B) whenever A c B and A, B E ~. The Choquet integral of a measurable function f with respect to a fuzzy measure p~ is defined by (C) f d / z =
foo /z({o~[ f ( t o ) > x } ) d x
+
[/x({oJ [ f ( ( o ) > x } ) - 1]dx. oo
where the integrals of the right side are ordinary ones. The following properties are well-known and can be easily shown.
Proposition 1.1 [2]. Let tz and v be fuzzy measures on ~. (1) For every measurable function f,
(c) f f d . =
x
where the integral of the right side is the Stieltjes integral and F(x) = -/x({oJ I f ( t o ) > x } )
Vxe(-~,
oo).
Correspondence to: Dr. T. Murofushi, Department of Communications and Systems, University of Electro-Communications, 1-5-1 Chofugaoka, Chofu, Tokyo 182, Japan. 0165-0114/93/$06.00 (~) 1993--Elsevier Science Publishers B.V. All rights reserved
T. Murofushi, M. Sugeno / Some quantities represented by the Choquet integral
230
(2) If v ~ IX, i.e., v(A ) <~ix(A) VA e ~, then, for every measurable function f,
(c)f f dv< (c)ff dix. 2. 0-1 fuzzy measures Definition 2.1. A 0-1 fuzzy measure is a fuzzy measure whose range is {0, 1}. Proposition 2.2. Let IX be a 0-1 fuzzy measure. For every measurable function f, ( C ) J( f dix =
sup A:/x(A)=I
inf f(o~). weA
Proof. Let a = SUpA:#(A)=I info~Af(~0). Taking into account Proposition 1.1(1), it is sufficient to prove that Ix({o~[f(o~)>x})=