Research journal of Applied Sciences, Engineering and Technology 4(8): 919-928, 2012 ISSN: 2040-7467 © Maxwell Scientific Organization, 2012 Submitted: October 15, 2011 Accepted: November 18, 2011 Published: April 15, 2012 The Banzhaf Value for Fuzzy Games with a Coalition Structure Fanyong Meng and Dong Tian School of Management, Qingdao Technological University, Qingdao, 266520, China Abstract: In this study, a general case of fuzzy games with a coalition structure is studied, where the player can participate in different unions. The expression of the Banzhaf value for this kind of fuzzy games is given, and its existence and uniqueness is shown. In order to better understand the Banzhaf value for this kind of fuzzy games, we discuss two special cases. The Banzhaf values on the associated fuzzy games are researched. Furthermore, some properties are studied. Finally, a numerical example is given. Key words: Banzhaf value, choquet integral, fuzzy game, multilinear extension Shapley function. Meng and Zhang (2010) researched the Shapley function for fuzzy games with multilinear extension form given by Owen (1971), and showed its existence and uniqueness. Li and Zhang (2009) studied the Shapley value for the general case of fuzzy games, which can be used in all kinds of fuzzy games. More researches refer to Sakawa and Nishizalzi (1994), Branzei et al. (2003), Tijs et al. (2004), Hwang (2007), Hwang and Liao (2008) and Butnariu and Kroupa (2009). In this study, we first propose the model of fuzzy games with a coalition structure, which is different to the coalition structure introduced by Aumann and Drèze (1974) and Owen (1977, 1978). The Banzhaf value for fuzzy games with a coalition structure is studied. Some properties are researched. Furthermore, we discuss two special kinds of fuzzy games with a coalition structure, and the associated Banzhaf values are given. Since the relationship between the fuzzy coalition values and that of their associated crisp coalitions is determined, their properties can be obtained by studying the associated games. INTRODUCTION As we know, in some cooperative games, not all coalitions can be formed. Aumann and Drèze (1974) gave a model of such games, where the players in a union are independent to other players. Different to Aumann and Drèze (1974), Owen (1977) researched games with a coalition structure where the probability of cooperation among unions is considered, and provided the Owen value, which is an extension of the Shapley value. The axiomatic systems of the Owen value are considered by Hamiache (1999), Khmelnitskaya and Yanovskaya (2007) and Albizuri (2008). Later, Owen (1978) proposed the Banzhaf- Owen value for games with a coalition structure, which is an extension of the Banzhaf value. The axiomatic systems of the Banzhaf- Owen coalition value is studied by Alonso-Meijide et al. (2007). Alonso-Meijde and Fiestras-Janeiro (2002) pointed the Banzhaf-Owen value dissatisfy symmetry in quotient game, and gave another solution concept for games with coalition structures, which is known as the symmetric coalitional Banzhaf value. On the other hand, there are some situations where some players do not fully participate in a coalition, but to a certain degree, this kind of games is called fuzzy games, which is first introduced by Aubin (1974). Owen (1971) defined a kind of fuzzy games, which are called fuzzy games with multilinear extension form. Tsurumi et al. (2001) defined a kind of fuzzy games with Choquet integral form, and the Shapley function defined on this class of games is given. Butnariu (1980) defined a class of fuzzy games with proportional value, and gave the expression of the Shapley function on this limited class of games. Recently, Butnariu and Kroupa (2008) expanded the fuzzy games with proportional value to fuzzy games with weighted function, and gave the corresponding PRELIMINARIES Let N = {1, 2, …, n} be a finite set, and P(N) denote the class of all subsets in N. The coalitions in P(N) are denoted by S0, T0, …. For any S0 , P(N), the cardinality of S0 is denoted by the corresponding lower case s. A coalition structure '={B10, B20,…,Bm0}on player set N is a partition of N, i.e., c1#h#m B h0 = N and Bh01Bl0 =i for all h,l0 M = {1, 2, ..., m} with h…l. A coalition structure on player set N is denoted by (N,'). For any S0 0 P(N,'). S0 is called a feasible coalition, where P(N,') denotes the set of all feasible coalitions in (N,'). A function v0 :P{N,'}6R+, such that v0 (i) = 0, is called a set function. The set of all set functions in (N,')is denoted by G0(N,'). Corresponding Author: Fanyong Meng, School of Management, Qingdao Technological University, Qingdao, 266520, China 919 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 Furthermore, when every union has the same cardinality, then Eq. (2) degenerates to be the Banzhaf value for fuzzy games on matroids. Owen (1978) proposed the Banzhaf-Owen value on G0(N,')as follows: i ( N , v0 , ) R M \ k i S0 B0k v0 ((Q0 S0 ) \ i ) 1 1 (v (Q S0 )) 2m1 2bk 1 0 0 i N Definition 1: Let v0 G U(M,B k ), v is said to be superadditive if: v(K) + v(T)#v(K wT),œ K T0 LU(M,Bk), (1) where Supp K SuppT = i. where Q0=cl,R Bl0. m and r denote the cardinalities of M and R, respectively. If there is no special explanation, for any v0 GU(M,Bk), we always mean v is superadditive. THE BANZHAF FUNCTION FOR FUZZY GAMES WITH A COALITION STRUCTURE Definition 2: Let v0 GU(M,Bk), the fuzzy coalition TfBk is said to be an f-carrier for v Bk in Bk, if we have v(KvT) = v(K) for any KfBk. Similar to Lehrer (1988), we give the concept of the “reduced” fuzzy game for v RBk in Bk. For any different indices i, j0 SuppBk, put R= {U(i), U(j)} and consider the “reduced” game VRBk in Bk, (with (Bk\R) v{R} as the set of players) defined by: In this section we shall research fuzzy games with a coalition structure. Different to the coalition structure given by Aumann and Drèze (1974) and Owen (1977, 1978), the coalition structure given in here does not required the intersection of different unions is empty set. A fuzzy coalition structure 'F = {B1, B2, ..., Bm} on fuzzy coalition U0L(N) is a set of unions on U, where w1#h#m Bh = U. Here, the players in a union are independent to other players, and the players can participate in different unions. A fuzzy coalition structure on player set U is denoted by (U,'F). Let P('F) denote the set of all probability distributions in 'F. For any P 0 P('F) and any Bk 0 'F, we have P(Bk) $0 and P( Bk ) 1 . Bk B B B v R k (T) = v k (T) and v R k (Tw{R}) = v (TwR) for any TfBk\R From above, we know the “reduced” fuzzy game v RBk has index |SuppBk|-1. Let f be a solution on v 0 GU(M,Bk). Similar to the crisp case, we give the following properties: Bk F By LU(M,Bk), we denote the set of all feasible coalitions in (U,'F) where M = {1, 2, …, m}. A function: : LU(M,Bk)6R+, such that v(i) = 0, is called a set function. The set of all set functions in LU(M,Bk) is denoted by GU(M,Bk). Let v0 GU(M,Bk) and P , P('F), we give the Banzhaf value on LU(M,Bk) as follows: i ( M , Bk , v ) P( Bk ) |Supp Bk | 1 k M S Bk ,iSuppS 2 v ( S )) i SuppU Probabilistic efficiency in unions (2-PEU): Let v 0 GU(M,Bk), P, P('F) and any Bk 0 'F, we have: f i ( Bk , v Bk ) f j ( Bk , LB k ) f R ( Bk , v RBk ) where i, j , SuppBk, and R = {U(i), U(j)}. ( v ( S U ( i )) (2) Null property (NP): Let v 0 GU(M,Bk) and P , P('F), if we have v (SvU(i)) = v (S) for any SfBk with I f Supp S and any k 0 M, then fi (M, Bk, v) = 0, where i0 SuppU. (3) Symmetry in unions (SU): Let v 0 GU(M,Bk) and P, P('F), if we have v (S w U(i)) = v (S w U(j) ) for any SfBk with if SuppS and any k 0 M, then we have fi(M,Bk,v) = 0, where i0 SuppU. From Eq. (2), we get: i ( M , Bk , v ) ( B ,v k M i Bk k ) i SuppU where i ( Bk , v Bk ) P( Bk ) Additivity (ADD): Let v1,v2 0 GU(M,Bk) and P , P('F), if we have: (v1+v2)(S) = v1(S)+ v2(S) for any S 0 LU(M,Bk), then: f(M,Bk,v1+v2) = f(M,Bk,v1) + f(M,Bk,v2) |SuppBk |1 s Bk 2 i SuppS ( v Bk ( SvU ( i )) v Bk ( S )) and v Bk denotes the restriction of v in Bk. From Eq. (2), we know when there is only one union in 'F, then Eq. (2) degenerates to be the Banzhaf value for traditional fuzzy games. Definition 3: Let v 0 GU(M,Bk) and P,P('F), the function f: GU(M,Bk)6{R+}|SuppU| is called a Banzhaf value for v in LU(M,Bk) if it satisfies 2-PEU, NP, SU and ADD. 920 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 Theorem 1: Let v 0 GU(M,Bk) and P,P('F), a value f satisfies 2-PEU, NP, SU and ADD if and only if f = n. From NP, we obtain f i ( Bk , v Bk ) P( Bk ) Proof: From Eq. (3), it is not difficult to get the existence. In the following, we shall prove the uniqueness holds. Let v 0 GU(M,Bk), we first show v can be expressed by: v P( B ) c u k k M cS |SuppS | P( Bk ) S Bk 2 1 f i ( Bk , v ) i SuppS 0 Bk where cS ( 1) |SuppS | |SuppT | v (T ) Namely, f(Bk, TS ST 1 u S (T ) 0 otherwise w U(i))- v Bk (S) = w Bk (S w U(i)) w Bk (S) for all SfBk with i 0 Supp Bk\Supp S, then we have: fi (Bk, v Bk ) = fj (Bk, v Bk ). Theorem 2: Let v 0 GU(M,Bk) and P , P('F) a value f satisfies 2-PEU, NP, SU and MU if and only if f = n. Proof: From Theorem 1 and Eq. (2), we know the existence holds. In the following, we will show the uniqueness. Define the index I of v to be the minimum number of non-zero terms in some expression for v of form (4). When Eq. (4) has only one fuzzy coalition i…TfBk such that cT…0, where kfM. From Theorem 1, we get. P( B k ) c S u S (T ) S Bk (T ) (5) From the uniqueness proof for traditional games, we get Eq. (5). Thus, Eq. (4) holds. From ADD, we only need to show the uniqueness of v in Bk for any k 0 M. Namely, the uniqueness of Eq. (3) for v Bk . Sinces: c u cT P( Bk ) |SuppT |1 i SuppT f i (U , cT uT ) 2 0 otherwise Thus, the result holds. Hypothesis, when I = m, we have the conclusion. In the following, we shall show the result holds when I = m+1. Without loss of generality, suppose: (6) S S S Bk We get: f i ( Bk , v Bk ) P( Bk ) B ) = n(Bk, v k ). GU(M,Bk), P , P('F), and any Bk 0 'F, if we have v Bk (S v (T ) P( Bk ) cS uS (T ) k M S Bk v Bk P( Bk ) otherwise Marginal contributions in unions (MU): Let v 0 For any T 0 LU (M,Bk), without loss of generality, suppose TfBk, then: cS uS S Bk v Bk i SuppS In the following, we give another axiom system of Eq. (2), which is inspired by Young (1985), Lehrer (1988) and Nowak (1997). and P( B k ) f i ( Bk , cS uS ) From 2-PEU and SU, we get: (4) S S S Bk S Bk m1 S Bk v cTq uT q 1 q f i ( Bk , cS uS ) 1 Let T = m q 1 Tq, for any i 0 SuppU\SuppT, construct the for any i 0 Supp Bk game: 921 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 for any i0SuppBk\ SuppTk. when i , SuppTk and |SuppTk| = |SuppBk | -1, without loss of generality, suppose SuppTkcj = Supp Bk. From Eq. (7), we have: w cTq uTq q:i Tq The index of wis at most I = m. Since: v(S w U(i))-v(S) = w(S w U(i))-w(S) i ( Bk , v B ) k for all S0 LU(M,Bk) with i ó Supp S. From MU and hypothesis, it follows that: f i (U , w) f i (U , v ) P( Bk ) k M q:i SuppTq 2 cT q |SuppTq |1 When i0 SuppT, amalgamate i with any other coalition j 0 SuppT, put R = {U(i), U(j)} and consider the “reduced” fuzzy game vR. From hypothesis, we get: f R (U , v R ) P( Bk ) k M qi Tq From 2-PEU and SU, we have: k M P( Bk ) k M cTq q:i SuppTq 2 |SuppTq |1 Tq U ,1 q m1 2 P ( Bk ) S Bk , i , j SuppS 2 |Supp Bk |1 ( S U (i )) v Bk ( S )) (v Bk ( S U (i )) v Bk ( S )) 2 P ( Bk ) S Bk ,i , j SuppS 2 |Supp Bk |1 k (v Bk ( S U (i )) v B ( S )) S Tk i SuppS S Tk i SuppS 2 P ( Bk ) |Supp Bk |2 2 P ( Bk ) Bk |Supp Tk |1 (v (v Bk ( S U (i )) v Bk ( S )) ( S U (i ) v Bk ( S )) i (Tk , v Bk ) cT q 2 |SuppTq |1 f i (U , v ) P( Bk ) S Bk , i , j Supps P ( Bk ) Bk |SuppBk |1 ( v v Bk ( S U (i ) U ( j )) v Bk ( S U ( j )) cTq 2 when i 0 SuppTk and |SuppTk| = |SuppBk| - p, without loss of generality, suppose Supp Tkc{j1, j2, …, jp}= SuppBk. Let T1=TkwU(k1), T2=T1 wU(k2),..., Tp= Tp-1 wU(kp). From above, we get: |SuppTq |1 Thus, we get f (U,v) = n (U,v). B ni(Tk, v k ) = ni (T1, v Bk ) =...= ni (Bk, v Bk ) Theorem 3: Let v 0 GU(M,Bk) and P , P('F), if TkfBk is Namely, n(Bk, v Bk ) = n(Tk, v Bk ) B an f-carrier for v k in Bk, where k0M, then: i ( M , Bk , v ) (T , v k M i k Bk ) i SuppU From Eq. (3), we get the conclusion. (7) TWO SPECIAL KINDS OF FUZZY GAMES WITH A COALITION STRUCTURE where i (Tk , v Bk ) S Tk i SuppS P ( Bk ) 2 |SuppS Tk |1 (v Bk ( S U (i )) v Bk ( S )) In this section, we shall research two special kinds of fuzzy games with a coalition structure, which are extensions of fuzzy games given by Owen (1971) and Tsurumi et al. (2001), respectively. Owen (1971) introduced fuzzy games with multilinear extension form. The fuzzy coalition value for this kind of fuzzy games is expressed by: Proof: Since Tk is an f-carrier for v Bk in Bk, we get: i ( Bk , v B ) i (Tk , v B ) 0 k k 922 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 vO (U ) From Eq. (10), we get: U (i ) (1 U (i )) v (T ) T0 SuppU i T0 0 i SuppU \ T 0 œU0L(N) i ( M , Bk , vO ) (8) where v0 is the associated crisp game with respect to vO. Let GoU (M,Bk) denote the set of all fuzzy games with multilinear extension form for GoU (M,Bk). Later, Tsurumi et al. (2001) proposed fuzzy games with Choquet integral form. The fuzzy coalition value for this kind of fuzzy games is written as: q (U ) v C (U ) v0 ([U ]hl )(hl hl 1 ) l 1 U L( N ) i SuppS œi0 Supp U i ( M , Bk , vO ) (9) P ( Bk ) U ( j) |SuppBk | 1 2 T0 j T0 j Supp ( S U ( i ))\ T0 k M S Bk Supp ( S U ( i )) i SuppS U ( j) j Supp ( S )\ T0 T0 Supp ( S ( i )) j T0 (10) Supp ( S U ( i )) j T0 (12) S Bk ,i SuppS j SuppS / T0 2 P ( Bk ) |SuppBk |1 (U (i ) U ( j ) T0 SuppS j T0 (1 U ( j ))) (v 0 (T0 i ) v 0 (T0 )) Proof:YFor any S 0 LU(M,Bk)such that ióSuppS, without loss of generality, suppose SfBk. When ió SuppBk, we easily get i is a null player for v0 in SuppBk. When i0 Supp Bk, we have: vo ( S U (i )) vO ( S ) ( U ( j) ) i SuppU Lemma 1: Let vO 0 GOU (M,Bk) , i is a fuzzy null player for vO 0 GOU (M,Bk) if and only if i is a null player for the associated crisp game v0 in SuppBk and any k0M. Since Bk O Definition 4: Let v0 be a crisp game defined in N, if we have v0 (S0 c i) = v0(S0) for any S0-fN, then i is said to be a null player for v0 in N. Similar to Definition 4, we can have the definition of fuzzy null player for v0 0 GOU (M,Bk) . Here, we omit. i T0 k when we restrict GU(M,Bk)in the setting of GOU(M,Bk), from Definition 3, we get the definition of the Banzhaf value on GOU(M,Bk) .Furthermore, from Theorem 1, we can show Eq. (10) is the unique Banzhaf value for v0 0 GOU(M, Bk). Here, we on longer discuss in detail. (1 U ( j )))v 0 (T0 ) i k i SuppU ( B ,v k M i ( Bk , vOB ) ni(M,Bk ,vo) (1 U ( j )))v0 (T0 )) (11) B where vo k denotes the restriction of vO in Bk, and The Banzhaf value for GOU(M,Bk): Let vO 0 GOU (M,Bk) and P-P('F), we give the Banzhaf value on LU(M,Bk)as follows: P( Bk ) U ( j) 2|SuppBk |1 i T0 Supp( S U (i )) j T0 and where [U]h1={i0N|U(i)$hl} and Q(U) ={U(i)|U(I)>0, i0 N}. q(U) denotes the cardinality of Q(U). The elements in Q(U) are written in the increasing order as 0 = h0 #h1 #…#h q(U). v0 is the associated crisp game with respect to vC. By GCU(M,Bk), we denote the set of all fuzzy games with Choquet integral form for GU(M,Bk). k M , S Bk (1 U ( j ))) v O ( S U (i )) v O ( S ) j Supp ( S U ( i ))\ T0 = (v0 (T0 ) v0 (T0 \ i )) 923 U ( j) (1 U ( j )) j Supp ( S U ( i ))\ T0 i T0 Supp ( S U ( i )) j T0 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 (v0 (T0 ) v0 (T0 \ i )) 0 i ( M , Bk , vO ) (T , v k M i k Bk O ) i SuppU (13) Since where U ( j) (1 U ( j )) 0 j Supp ( S U ( i ))\ T0 i T0 Supp ( S U ( i )) j T0 i (Tk , v Bk ) U (i ) U ( j ) P ( Bk ) j T0 |SuppTk |1 T0 SuppS (1 U ( j )) S Tk 2 i SuppS j SuppS \T0 we have: v0 (T0 ) v0 (T0 \ i ) 0 T0 Supp( S U (i )) (v 0 (T0 i ) v 0 (T0 )) where i0 T0. From the randomicity of S. we know i is a null player for v0 in SuppBk and any k 0 M. Z Proof: For any i0 SuppBk\SuppTk, we know i is a null player for the associated crisp game v0 in SuppBk. From Lemma 1, we have i is a fuzzy null player for voBk . To prove i is a fuzzy null player for vO 0 GOU (M,Bk), we only need to show i is a fuzzy null player for voBk and any k0M. When ió SuppBk, it is obviously Thus,ni (Bk , v k ) = 0. When i 0 SuppTk and |SuppTk| = |SuppBk | -1, without loss of generality, suppose SuppTkcj = SuppBk. From Eq. (12), we have: B that i is a fuzzy null player for vo k . When i0 SuppBk, for any SfBk with i ó SuppS, we have: B i ( Bk , vo k ) B vO ( S U (i )) vO ( S ) S Bk (1 U ( j )) U ( j) i T0 Supp ( S U ( i )) j T0 j Supp ( S U ( i ))\T0 i SuppS P( Bk ) U (i ) U ( p) (1 U ( p) pT0 pSuppS \T0 2 |SuppBk |1 T0 SuppS (v0 (T0 i ) v0 (T0 )) (v0 (T0 ) v0 (T0 \ i )) Since v0(T0)-v0(T0\I) = 0, we have: vO(SwU(i))-vO(S) = 0 S Bk i , j SuppS P( Bk ) (1 U ( p)) U (i ) U ( p) 2 |SuppBk |1 T0 SuppS p T0 pSuppS \ T0 Thus, i is a fuzzy null player for voBk Lemma 2: Let vO 0 G (M,Bk), Tk 0 Bkis an f- carrier for v if and only if SuppTkfSuppBk is a carrier for the associated crisp game v0 in SuppBk. O U Bk o T0 SuppS (1 U ( p)) U (i )1 U ( j )) U ( p) pT0 pSuppS \ T0 (v0 (T0 i ) v0 (T0 )) Proof: From the definitions of the f-carrier for fuzzy games and the carrier for crisp games, we know it is sufficiency to show i is a fuzzy null player for v Bk if and only if i is a null player for the associated crisp game v0 in SuppBk. From Lemma 1, we get the conclusion. Theorem 4: Let vO 0 GOU (M,Bk) and P0P('F), if SuppTkfSuppBk is a carrier for the associated crisp game v0 in SuppBk, where k0M. Then, p SuppS / To S Bk i . j SuppS 924 T0 SuppS (U (i )U ( j ) U ( p)) pT0 (1 U ( p))(v0 (T0 {i , j}) v0 (T0 j )) P( Bk ) U (i ) U ( p) (1 U ( p)) 2 |SuppBk |1 T0 SuppS pT0 pSuppS \ T0 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 (v0 (T0 i ) v0 (T0 )) i ( Bk , vOB ) U (i )(1 U ( j )) U ( p) (1 U ( p)) p T0 pSuppS \ T0 T0 SuppS k U (i )U ( j ) U ( p) pT0 T0 SuppS S Bk i SuppS (1 U ( p)) (v 0 (T0 i ) v 0 (T0 )) pSuppS \ T0 U ( j ))))(v0 (T0 i ) v0 (T0 )) S Bk i , j SuppS U ( j ))))(v0 (i ) S Tk i SuppS 2 P ( Bk ) |SuppTk |1 (U (i )) U ( p) T 0 SuppS pT0 S Bk ,i SuppS 2 P ( Bk ) |Supp Bk |1 Thus, we have: i ( M , Bk , vO ) = i (Tk , vO k ) . B i ( Bk , v O k ) B When i0SuppTk and |SuppTk| = |SuppBk| - q, without loss of generality, suppose SuppTkc{j1, j2,…, jq} = SuppBk. Let T1 = Tkw U(k1),T2=T1 wU(k2),...,Tq=Tq-1wU(kq). From above, we get: k M k k k i SuppS 2 |SuppBk |1 q(S) (v0 [ S ]hl ) v0 ([ S \ U (i )]hl ) l 1 œi 0SuppS (14) From Eq. (14), we further have: i ( M , Bk , vC ) Proof: From the superadditivity of v0, we have v0(S0 c i)v0(S0 )$v0(i) for any S0fN with ióS0. U (i ) U ( j ) (1 U ( j )) U (i ) j T0 j SuppS \ T0 P( Bk ) ×(hl-hl-1) Theorem 5:Let vO 0 GOU (M,Bk) and P0P('F), if the associated crisp game v0 is superadditive, then we have ni (M,Bk,vO)for any i0 SuppU. k M S Bk k M The result is obtained. T0 SuppS P( Bk )v0 (i )U (i ) i ( M , Bk , vC ) Thus, i ( M , Bk , vO ) i ( Bk , vOBk ) i (Tk , vOBk ) k M k M ,i SuppBk The Banzhaf value for: GCU(M,Bk): Let Vc0GCU (M,Bk) and P0P('F), we give the Banzhaf value onLU(M,Bk)as follows: i (Tk , vOB ) i (T1 , vOB ) ,... i ( Bk , vOB ) . Since, U (i )v 0 (i ) = P(Bk)U(i)v0 (i) U ( p))))(v 0 (T0 i ) v 0 (T0 )) pSuppS \T0 P( Bk ) (U (i ) (U ( j ) (1 |SuppBk |1 T0 SuppS j T0 j SuppS \ T0 2 S Bk i SuppS 2 P( Bk ) 2 |SuppBk |1 T0 SuppS U (i ) (1 U ( p) (1 U ( p))))(v0 (T 0 i ) v0 (T0 )) pT0 pSuppS \ T0 P( Bk ) (U (i ) U ( j ) (1 j T0 j SuppS \ T0 2 |SuppBk |1 T0 SuppS q ( Bk ) P( B ) ([ B ] k M k l 1 i k hl , v0 )(hl hl 1 ) (15) where for i ([ Bk ]h , v0 ) l S0 [ Bk ] hl 1 2 |[ Bk ]hl | 1 (v0 ( S0 ) v0 ( S0 \ i )) iS0 any SfBk. From Eq. (12), we get: and|[Bk]hl| denotes the cardinality of [Bk]hl. 925 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 Theorem 7: Let vC 0 GCU (M,Bk) and P0P('F), if TkfBk is an f-carrier for vC in Bk, then Similar to GOU (M,Bk), when we restrict GU (M,Bk) in the setting of GCU (M,Bk), from Definition 3, we get the definition of the Banzhaf value on GCU (M,Bk). Furthermore, from Theorem 1, we can show Eq. (14) is the unique Banzhaf value for vC 0 GCU (M,Bk). Here, we also on longer discuss in detail. i ( M , Bk , vC ) Theorem 6: Let vc0 GCU (M,Bk) and any i,j 0 SuppBk, then vC(S wU(i)) = vC(S wU(j)) for any SfBk with i,j ó SuppS if and only of v0(S0 c i) = v0 (S0 c j) for any S0f[Bk]hl with i,j óS0 and i,j 0[Bk]hl wherev0 is the associated crisp game of vC and #l #q (Bk). q ( Tk ) P( B ) ([T ] k k M l 1 i k hl , v0 )(hl hl 1 ) i SuppU (16) where i ([Tk ]h , v0 ) l Proof: From Eq. (9), it is not difficult to get the result. S0 [ Tk ]he ,i S0 Lemma 3: Let vc0 GCU (M,Bk) is an f- carrier for vC in Bk if and only if [Tk]klf[Bk]he is a carrier for the associated crips game v0 in [Bk]he for any 1#l #q(Bk). = ni ([Tk]hl,v0) l1 1 vC(SwU(i)) (v0 ( S0 ) v0 ( S0 \ i )) ni ([Bk]hl,v0) YFor any SfBk with i ó SuppS, since: q ( S U ( i )) 2 |[ Tk ]he | 1 Proof: From Theorem 6, we know Tk]kl f [Bk] is a carrier for v0 in [Bk]kl. Similar to Theorem 4, we get: Proof: In order to show the conclusion, we only need to prove i is a fuzzy null player for vC in Bk if and only if i is a null player for v0 in [Bk] and any1#l #q (Bk) . 1 S0 [ Tk ]kl ,i S0 2 |[ Tk ]he | 1 (v0 ( S0 ) v0 ( S0 \ i )) i [ Bk ]hl v0 ([ S U (i )]hl )(hl hl 1 ) Thus, q ( Bk ) ([ B q ( S U ( i )) q (U (i )) v0 ([ S ]hl i ) v0 ([ S ]hl ) (hl hl 1 ) l1 l 1(U ( i )) 1 l 1 i ] , v0 )(hl hl 1 ) k hl q ( Bk ) ([T ] l 1 i k hl , v 0 )(hl hl 1 ) and Since vC (SwU(i)) = vC(S) we get (v ([ S ] l 1 0 hl q ( Bk ) q ( Tk ) l 1 l 1 ) i ([Tk ]hl , v0 )(hl hl 1 ) i ([Tk ]hl , v0 )(hl hl 1 ) q (U ( i )) i ) v0 ([ S ]hl )(hl hl 1 ) 0 . We get the conclusion. Theorem 8: Let vC 0 GCU (M,Bk) and P0P('F),if the associated crisp game v0 is superadditive, then we have i ( M , Bk , v c ) P( Bk )v0 (i)U (i) for any i0 SuppU. Namely, v0([S]hl c i) = v0 ([S]hl c i)-for any1#l #q (U(i)). From the randomcity of S, we get i is a null player for v0 in [Bk]hl , where1#l #q (U(I)). Furthermore, it is obvious that i is a null player for v0 in [Bk]hl , where q (U(i)) <l #q (U(Bk)). Z If i is a null player for v0 in[Bk]kl for any1#l#q (U(Bk)), from Eq.(9), we know i is a fuzzy null player for vC in Bk. k M , i SuppBk Proof: The proof of Theorem 8 is similar to that of Theorem 5. 926 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 Table 1: The crisp coalition values v0(S0) S0 {1} 2 {2} 2 {3} 1 {4} 2 {5} 1 {1,2} 6 {1,3} 6 {1,4} 5 {2,3} 5 {2,4} 7 S0 {3,4} {2,5} {4,5} {1,2,3} {1,2,4} {1,3,4} {2,3,4} {2,4,5} {12,3,4} learning scope of game theory. The Banzhaf value for this kind of fuzzy games is studied. In order to better understand this class of fuzzy games, we further study two special cases, which can be seen as the extensions of fuzzy games given by Owen (1971) and Tsurumi et al. (2001). v0(S0) 6 5 6 15 18 16 16 18 30 ACKNOWLEDGMENT This study was supported by the National Natural Science Foundation of China (Nos 70771010, 70801064 and 71071018). NUMERICAL EXAMPLE There are 5 companies (which represent the players 1, 2, 3, 4 and 5), who will possibly cooperate. As a result of the differences in technical and fund among them, the players 1, 2, 3 and 4 decide to cooperate and the players 2, 4 and 5 decide to cooperate. As is in the real life, each company is not willing to supply all its resources to a particular cooperation. Thus, we have to consider a fuzzy game. Consider a fuzzy coalition U defined by: REFERENCES Albizuri, M.J., 2008. Axiomatizations of the Owen value without efficiency. Math. Soc. Sci., 55(1): 78-89. Alonso-Meijde, J.M. and M.G. Fiestras-Janeiro, 2002. Modification of the Banzhaf value for games with a coalition structure. Ann. Oper. Res., 109 (1-4): 213-227. Alonso-Meijide, J.M., F. Carreras, M.G. Fiestras-Janeiro and G. Owen, 2007. A comparative axiomatic characterization of the Banzhaf-Owen coalitional value. Decis. Support Syst., 43(3): 701-712. Aubin, J.P., 1974. Coeur et valeur des jeux flous à paiements latéraux. Comptes Rendus Hebdomadaires D., 279-A: 891-894. Aumann, R.J. and J.H. Drèze, 1974. Cooperative games with coalition structures. Int. J. Game Theory, 3(4): 217-237. Banzhaf, J., 1965. Weighted voting does not work: A mathematical analysis. Rutgers Law Rev., 19: 317-343. Branzei, R., D. Dimitrov and S. Tijs, 2003. Convex fuzzy games and participation monotonic allocation schemes. Fuzzy Set Syst., 139(3): 267-281. Butnariu, D., 1980. Stability and Shapley value for an npersons fuzzy game. Fuzzy Set. Syst., 4(1): 63-72. Butnariu, D. and T. Kroupa, 2008. Shapley mappings and the cumulative value for n-person game with fuzzy coalitions. Eur. J. Oper. Res., 186(1): 288-299. Butnariu, D. and T. Kroupa, 2009. Enlarged cores and bargaining schemes in games with fuzzy coalitions. Fuzzy Set Syst., 160(5): 635-643. Hamiache, G., 1999. A new axiomatization of the Owen value for games with coalition structures. Math Soc. Sci., 37(3): 281-305. Hwang, Y.A., 2007. Fuzzy games: A characterization of the core, Fuzzy Set. Syst., 158(22): 2480-2493. Hwang, Y.A. and Y.H. Liao, 2008. Max-consistency, complement-consistency and the core of fuzzy games. Fuzzy Set. Syst., 159(2): 152-163. U(1) = 0.6, U(2) = 0.3, U(3) = 0.8 U(4) = 0.7, U(5) = 0.5 From above, we know N = {1, 2, 3, 4, 5}, B1 ={U(1), U(2), U(3), U(4)}and B2 ={U(2), U(4), U(5)}, and this is a fuzzy game belongs to GU(M,Bk ). The crisp coalition values are given by Table 1. If we use the occupying proportions of the associated crisp coalition values of B1 and B2 as the probability distribution, then we obtain P(B1) = 3/8 and P(B2) = 5/8. When the fuzzy coalition values and their associated crisp coalition values have the relationship given by Eq. (8). Namely, this fuzzy game belongs to GOU(M,Bk). From Eq. (10) , (12), we obtain the player Banzhaf values are n1 (M,Bk,v0) = 1066, n2(M,Bk,v0) = 1.6206 n3 (M,Bk,v0) = 1.1603, n4(M,Bk,v0) = 2.5424 n5 (M,Bk,v0) = 0.793 When the fuzzy coalition values and their associated crisp coalition values have the relationship given by Eq. (9). Namely, this fuzzy game belongs to GcU (M,Bk). From Eq. (14), (15), we obtain the player Banzhaf values are n1 (M,Bk,vC) = 1.3922, n2(M,Bk,vC) = 1.9359 n3 (M,Bk,vC) = 1.4297, n4(M,Bk,vC) = 3.4391,n5 (M,Bk,vC) = 1.2031 CONCLUSION We have researched a more general case of cooperative fuzzy games, which allows the player to participate in different unions. The study extends the 927 Res. J. Appl. Sci. Eng. Technol., 4(8): 919-928, 2012 Khmelnitskaya, A.B. and E.B. Yanovskaya, 2007. Owen coalitional value without additivity axiom. Math Method Oper. Res., 66(2): 255-261. Lehrer, E., 1988. An axiomatization of the Banzhaf value. Int. J. Game Theory, 17(2): 89-99. Li, S.J. and Q. Zhang, 2009. A simplified expression of the Shapley function for fuzzy game. Eur. J. Oper. Res., 196(1): 234-245. Meng, F.Y. and Q. Zhang, 2010. The Shapley function for fuzzy cooperative games with multilinear extension form. Appl. Math. Lett., 23(5): 644-650. Nowak, A.S.,1997. On an axiomatization of the Banzahf value without the additivity axiom. Int. J. Game Theory, 26(1): 137-141. Owen, G., 1971. Multilinear extensions of games. Manage. Sci., 18(5): 64-79. Owen, G., 1977. Values of Games with a Priori Unions. Springer-Verlag, Nueva York, pp: 76-88. Owen, G., 1971. Multilinear extensions of games. Manage. Sci., 18(5): 64-79. Owen, G., 1977. Values of Games with a Priori Unions. Springer-Verlag, Nueva York, pp: 76-88. Owen, G., 1978. Characterization of the banzhaf-coleman index. SIAM J. Appl. Math., 35(2): 315-327. Sakawa, M. and I. Nishizalzi, 1994. A lexico-graphical solution concept in an n-person cooperative fuzzy game. Fuzzy Set. Syst., 61(3): 265-275. Tsurumi, M., T. Tanino and M. Inuiguchi, 2001. A Shapley function on a class of cooperative fuzzy games. Eur. J. Oper. Res., 129(3): 596-618. Tijs, S., R. Branzei, S. Ishihara, and S. Muto, 2004. On cores and stable sets for fuzzy games. Fuzzy Set Syst., 146(2): 285-296. Young, H., 1985. Monotonic solutions of cooperative games. Int. J. Game Theory, 14(2): 65-72. 928