Interaction in Choquet inegral model
This paper studies the notion of interaction between criteria in a Choquet integral model.
This paper studies the notion of interaction between criteria in a Choquet integral model.
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Then µ(K ∪ L) = (2n) q = (2n)(2n) q−1 and µ(K ∪ L \ {i}) = (2n) r i
.
So we have ∑ µ(K ∪ L \ {i}) = ∑ (2n) r i
≤ ∑
i∈L
i∈L
i∈L(2n) q−1 = l(2n) q−1 .
As 2n > l, then µ(K ∪ L) > ∑ µ(K ∪ L \ {i}) , hence we have
i∈L
∑
µ(K ∪ L) > ∑ ∑
µ(K ∪ L \ {i}).
L⊆A,
L⊆A, i∈L
l=a−p
l=a−p
According to Remark 4 (with t = a − p), we have
∑ ∑
a∑
µ(K∪L\{i}) = (p+1) µ(K∪L) ≥
L⊆A,
l=a−p
i∈L
Then ∑ L⊆A,
l=a−p
Therefore
µ(K ∪ L) >
a∑
p=0,
p even
[ ∑
L⊆A,
l=a−p
a∑
L⊆A,
l=a−p−1
L⊆A,
l=a−p−1
µ(K ∪ L) −
µ(K ∪ L), i.e., [ ∑
∑
L⊆A,
l=a−p−1
L⊆A,
l=a−p
a∑
L⊆A,
l=a−p−1
]
µ(K ∪ L)
µ(K∪L) since ( 1
p+1)
= p+1 ≥ 1.
µ(K ∪ L) −
a∑
L⊆A,
l=a−p−1
> 0, i.e., ∆ µ(K)
A
> 0.
The following example illustrates this two previous results in this section.
Example 2. N = {1, 2, 3, 4}, P = {(a 23 , a 1 ), (a 234 , a 123 ), (a 2 , a 13 )}.
µ(K ∪ L) ] > 0.
The ordinal preference information {P, I} contains no indifference and the binary relation
(P ∪ M) contains no strict cycle, so {P, I} is representable by a general Choquet integral
model. A suitable topological sorting on (P ∪M) is given by: B 0 = {a 0 }; B 1 = {a 1 , a 3 , a 4 };
B 2 = {a 13 , a 14 , a 34 }; B 3 = {a 2 }; B 4 = {a 12 , a 23 , a 24 }; B 5 = {a 123 , a 124 , a 134 }; B 6 = {a 234 }
and B 7 = {a N }. The Table 5 gives a capacity µ compatible with preference information
{P, I} and corresponding interaction indices.
S 8 7 × µ(S) I µ S
{1}, {3} 8 −
{2}, {4} 8 3 −
{1, 2} 8 4 0.33
{1, 3}, {1, 4} 8 2 0.33
{2, 3}, {2, 4} 8 4 0.33
{3, 4} 8 2 0.33
{1, 2, 3}, {1, 2, 4}, {1, 3, 4} 8 5 0.49
{2, 3, 4} 8 6 0.50
N 8 7 0.99
Table 5: A capacity µ ∈ C Pref and the corresponding interaction indices.
We can see that I µ A
> 0, ∀A ⊆ N such that |A| ≥ 2.
13