Next Article in Journal
Detecting and Handling Cyber-Attacks in Model Predictive Control of Chemical Processes
Next Article in Special Issue
On Metric Dimensions of Symmetric Graphs Obtained by Rooted Product
Previous Article in Journal
Almost Periodic Solutions of First-Order Ordinary Differential Equations
Previous Article in Special Issue
Computing The Irregularity Strength of Planar Graphs
Article Menu
Issue 9 (September) cover image

Export Article

Mathematics 2018, 6(9), 172; doi:10.3390/math6090172

Article
Optimizing Three-Dimensional Constrained Ordered Weighted Averaging Aggregation Problem with Bounded Variables
Department of Industrial Engineering and Management, National Kaohsiung University of Science and Technology, Kaohsiung 80778, Taiwan
*
Authors to whom correspondence should be addressed.
Received: 3 September 2018 / Accepted: 14 September 2018 / Published: 19 September 2018

Abstract

:
A single constrained ordered weighted averaging aggregation (COWA) problem is of considerable importance in many disciplines. Two models are considered: the maximization COWA problem with lower bounded variables and the minimization COWA problem with upper bounded variables. For a three-dimensional case of these models, we present the explicitly optimal solutions theoretically and empirically. The bounds and weights can affect the optimal solution of the three-dimensional COWA problem with bounded variables.
Keywords:
constrained ordered weighted averaging aggregation problem; mixed integer linear programming; bounded variables

1. Introduction

An ordered weighted averaging (OWA) operator [1] is a general class of parametric aggregation operators that appears in many research fields such as decision making [2,3,4,5,6], fuzzy system [7,8], statistics [9,10,11], risk analysis [12] and others [13,14]. For more details, see Carlsson and Fullér [15], Emrouznejad and Marra [16] and Yager et al. [17]. A constrained OWA aggregation problem (COWA) attempts to optimize the OWA aggregation problem with multiple constraints. Yager [18] developed a mixed integer linear programming problem to solve a single COWA problem. Later, Carlsson et al. [19] proposed an algorithm to solve the single constrained OWA optimization problem for any dimensions. Furthermore, Coroianu and Fullér [20] presented an explicitly optimal solution for the single COWA problem with any constraint coefficients. In addition, there are other important references [21,22,23,24,25,26,27] dedicated to constrained OWA optimization problem with multiple constraints. However, the decision variables are usually bounded for the most practical problems. Recently, Chen and Tang [28] proposed a three-dimensional COWA problem with lower bounded variables. This paper presents the explicitly optimal solutions for the three-dimensional COWA problem with bounded variables. Two models are considered. One is a maximizing three-dimensional constrained OWA aggregation problem with lower bounded variables (3COWAL). The other is a minimizing three-dimensional constrained OWA aggregation problem with upper bounded variables (3COWAU).
The organization of this paper is as follows. Section 2 briefly reviews the COWA problem. For maximizing 3COWAL, there are two parameters ( w 1 , w 2 , w 3 ) and ( l 1 , l 2 , l 3 ) that affect the optimal solution types. We discuss the optimal solution behaviors in Section 3 for w 1 w 2 w 3 and Section 4 for l 1 l 2 l 3 . Section 5 analyzes the optimal solution behaviors of minimizing 3COWAU. Finally, some concluding remarks are presented.

2. Constrained OWA Aggregation Problem

An OWA operator of dimension n is a mapping F : R n R that has an associated weighting vector W = ( w 1 , w 2 , , w n ) satisfying
w 1 + w 2 + + w n = 1 ,   0 w i 1 ,   i = 1 , 2 , , n
and such that
G ( x 1 , x 2 , , x n ) = i = 1 n w i y i
with y i being the ith largest of { x 1 , x 2 , , x n } . For simplicity, we will denote this expression as F ( y 1 , y 2 , , y n ) .
Consider the following single COWA problem:
Max   W T Y ,   s . t .   I T X 1 ,   X 0 ,
where the column vectors X, Y, W and I are
X = [ x 1 x 2 x n ] ,   Y = [ y 1 y 2 y n ] ,   W = [ w 1 w 2 w n ] ,   I = [ 1 1 1 ] .
By introducing the ( n 1 ) × n matrix
G = [ 1 1 0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 1 ]
and the column binary vectors Z i { 0 , 1 } n , i = 1 , 2 , , n , Yager [18] transformed the nonlinear programming single COWA problem (2) to the following mixed integer linear programming problem (MILP):
Max   W T Y ,   s . t .   I T X 1 ,   G Y 0 ,   y i I X M Z i 0 ,   i = 1 ,   2 , , n 1 ,   y n I X 0 ,   I T Z i n i ,   i = 1 ,   2 , , n 1 ,   Z i { 0 , 1 } n ,   i = 1 ,   2 , , n 1 ,   X 0 ,
where M is a very large positive number. To reduce the multiple solutions of the MILP (3), Chen and Tang [28] introduced the following constraints:
Z i + 1 Z i ,   i = 1 ,   2 , , n 2 .
Then, the more efficient MILP of a single COWA problem is as follows:
Max   W T Y ,   s . t .   I T X 1 ,   G Y 0 ,   y i I X M Z i 0 ,   i = 1 ,   2 , , n 1 ,   y n I X 0 ,   I T Z i n i ,   i = 1 ,   2 , , n 1 ,   Z i + 1 Z i ,   i = 1 ,   2 , , n 2 ,   Z i { 0 , 1 } n ,   i = 1 ,   2 , , n 1 ,   X 0 .

3. Maximizing a Three-Dimensional Constrained OWA Aggregation Problem with Lower Bounded Variables for w1w2w3

It is fairly common in practical optimization problems that the decision variables are usually bounded. A typical decision variable x i is bounded from below by l i and from above by u i , where l i u i . Section 3 and Section 4 analyze the maximization COWA problem with the lower bound constraints. Section 5 analyzes the minimization COWA problem with the upper bound constraints. Chen and Tang [28] proposed the following COWAL:
Max   W T Y ,   s . t .   I T X 1 ,   G Y 0 ,   ,   y i I X M Z i 0 ,   i = 1 ,   2 , , n 1 ,   y n I X 0 ,   I T Z i n i ,   i = 1 ,   2 , , n 1 ,   Z i + 1 Z i ,   i = 1 ,   2 , , n 2 ,   Z i { 0 , 1 } n ,   i = 1 ,   2 , , n 1 ,   X L ,
where the lower bounded vector
L = [ l 1 l 2 l n ] .
The lower bounded vector can be transformed into the zero vector by using the standard transformations X = X L . The COWAL is as follows:
Max   W T Y ,   s . t .   I T X 1 I T L ,   G Y 0 ,   y i I X M Z i L ,   i = 1 ,   2 , , n 1 ,   y n I X L ,   I T Z i n i ,   i = 1 ,   2 , , n 1 ,   Z i + 1 Z i ,   i = 1 ,   2 , , n 2 ,   Z i { 0 , 1 } n ,   i = 1 ,   2 , , n 1 ,   X 0 .
If 1 I T L < 0 , then the COWAL has no feasible solution. If 1 I T L = 0 , then X = 0 is the unique optimal solution, so X = L . The following will discuss the case that 1 I T L > 0 .
Consider the 3COWAL for the case of
l 1 + l 2 + l 3 1 .
Two parameters ( w 1 , w 2 , w 3 ) and ( l 1 , l 2 , l 3 ) are considered in 3COWAL. This section discusses 3COWAL with
w 1 w 2 w 3 .
There are six permutations of ( l 1 , l 2 , l 3 ) . First, consider the case of
l 1 l 2 l 3 .
At optimality, the first constraint of model (6) becomes
x 1 + x 2 + x 3 = 1 l 1 l 2 l 3 .
There are three types (A, B, C) of ( x 1 , x 2 , x 3 ) according to the number of zero components.
The number of zero components of first type A is two. The possible values of ( x 1 , x 2 , x 3 ) are
( 1 l 1 l 2 l 3 , 0 ,   0 ) ,   ( 0 , 1 l 1 l 2 l 3 , 0 )   and   ( 0 ,   0 ,   1 l 1 l 2 l 3 ) .
For the case of ( x 1 , x 2 , x 3 ) = ( 1 l 1 l 2 l 3 , 0 ,   0 ) , we have
( x 1 , x 2 , x 3 ) = ( 1 l 2 l 3 , l 2 , l 3 )
and
( y 1 , y 2 , y 3 ) = ( 1 l 2 l 3 , l 2 , l 3 ) ,   ( l 2 , 1 l 2 l 3 , l 3 ) ,   ( l 2 , l 3 , 1 l 2 l 3 ) ,   ( 1 l 2 l 3 , l 3 , l 2 ) ,   ( l 3 , 1 l 2 l 3 , l 2 )   or   ( l 3 , l 2 , 1 l 2 l 3 ) .
Consider ( y 1 , y 2 , y 3 ) = ( l 2 , 1 l 2 l 3 , l 3 ) , so l 2 1 l 2 l 3 l 3 , implying 2 l 2 + l 3 1 . From Label (7), it follows that l 2 l 1 , in contradiction to assumption (9). The same contradiction is also for ( y 1 , y 2 , y 3 ) = ( l 2 , l 3 , 1 l 2 l 3 ) , ( 1 l 2 l 3 , l 3 , l 2 ) , ( l 3 , 1 l 2 l 3 , l 2 ) and ( l 3 , l 2 , 1 l 2 l 3 ) . Therefore, for ( x 1 , x 2 , x 3 ) = ( 1 l 1 l 2 l 3 , 0 ,   0 ) , the reasonable candidate for the optimal solution is ( y 1 , y 2 , y 3 ) = ( 1 l 2 l 3 , l 2 , l 3 ) . For cases of ( 0 ,   1 l 1 l 2 l 3 , 0 ) and ( 0 ,   0 ,   1 l 1 l 2 l 3 ) , the reasonable candidates of ( y 1 , y 2 , y 3 ) are shown in Table 1.
In Table 1, there are six candidates for optimal solution ( y 1 , y 2 , y 3 ) for type A. Among these six candidates, we will show that the largest objective function F ( Y ) = w 1 y 1 + w 2 y 2 + w 3 y 3 is that of ( y 1 , y 2 , y 3 ) = ( 1 l 2 l 3 , l 2 , l 3 ) . Before we prove this result in detail, we present a well-known fact.
Theorem 1.
For ( x 1 , x 2 , x n ) , ( x 1 , x 2 , x n ) , s k = i = 1 k x i and s k = i = 1 k x i , k = 1 , 2 , , n , if s k s k , k = 1 , 2 , , n ,   then for all ( w 1 , w 2 , , w n ) with w k w k + 1 , k = 1 , 2 , , n 1 , we have
i = 1 n w i x i i = 1 n w i x i .
Comparing the objective function value of ( y 1 , y 2 , y 3 ) = ( 1 l 2 l 3 , l 2 , l 3 ) with that of ( 1 l 1 l 3 , l 1 , l 3 ) , we get that
  s 1 = 1 l 2 l 3 s 1 = 1 l 1 l 3 ,
s 2 = 1 l 3 s 2 = 1 l 3 ,
s 3 = 1 s 3 = 1 .
It implies that the most favorable value of the objective function is that with ( 1 l 2 l 3 , l 2 , l 3 ) . A similar argument shows that F ( 1 l 2 l 3 , l 2 , l 3 ) is larger than those of ( l 1 , 1 l 1 l 3 , l 3 ) , ( 1 l 1 l 2 , l 1 , l 2 ) , ( l 1 , 1 l 1 l 2 , l 2 ) and ( l 1 , l 2 , 1 l 1 l 2 ) . Therefore, the optimal solution for type A is ( 1 l 2 l 3 , l 2 , l 3 ) .
For the one zero component of ( x 1 , x 2 , x 3 ) , the possible values are
( x 1 , x 2 , 0 ) ,   ( x 1 , 0 , x 3 )   and   ( 0 , x 2 , x 3 ) .
At optimal, the possible values of x 1 , x 2 and x 3 with at least one x i = 0 , i = 1 ,   2 ,   3 satisfy
x 1 + l 1 = x 2 + l 2 ,   x 1 + l 1 = x 3 + l 3   or   x 2 + l 2 = x 3 + l 3 .
We choose x i + l i = x j + l j for type B1, and x i + l i = l k or x j + l j = l k , i   j   k , i ,   j ,   k = 1 ,   2 ,   3 , for type B2. A similar argument shows that all possible candidates for optimal solutions ( y 1 , y 2 , y 3 ) are
( 1 l 3 2 , 1 l 3 2 , l 3 ) ,   ( 1 l 2 2 , 1 l 2 2 , l 2 ) ,   ( l 1 , 1 l 1 2 , 1 l 1 2 )
for type B1, and
( 1 2 l 2 , l 2 , l 2 ) ,   ( l 1 , l 1 , 1 2 l 1 )
for type B2. The largest objective function value is that with ( 1 l 3 2 , 1 l 3 2 , l 3 ) for type B1 and ( l 1 , l 1 , 1 2 l 1 ) for type B2. Furthermore, F ( 1 l 3 2 , 1 l 3 2 , l 3 ) F ( l 1 , l 1 , 1 2 l 1 ) . Therefore, the optimal solution for type B is ( 1 l 3 2 , 1 l 3 2 , l 3 ) .
Type C is the nonzero components. From Label (10), it follows that
( x 1 , x 2 , x 3 ) = ( 1 / 3 l 1 , 1 / 3 l 2 , 1 / 3 l 3 ) ,   ( x 1 , x 2 , x 3 ) = ( 1 / 3 , 1 / 3 , 1 / 3 )   and   ( y 1 , y 2 , y 3 ) = ( 1 / 3 , 1 / 3 , 1 / 3 ) .
Therefore, there are six candidate optimal solutions for type A, five candidate optimal solutions for type B and one candidate optimal solution for type C. Detailed results of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) , ( y 1 , y 2 , y 3 ) , F ( y 1 , y 2 , y 3 ) and condition for 3COWAL with l 1 l 2 l 3 are presented in Table 2.
The largest objective function value is that of A 1 ( 1 l 2 l 3 , l 2 , l 3 ) for type A, B 1 ( 1 l 3 2 , 1 l 3 2 , l 3 ) for type B and C ( 1 / 3 , 1 / 3 , 1 / 3 ) for type C. A similar argument shows that
F ( A 1 ) F ( B 1 )   and   F ( A 1 ) F ( C ) .
Therefore, for the case of w 1 w 2 w 3 , the optimal solution for l 1 l 2 l 3 is A 1 . Similarly, the optimal solutions of the remaining five permutations l 1 l 3 l 2 , l 2 l 1 l 3 , l 2 l 3 l 1 , l 3 l 1 l 2 and l 3 l 2 l 1 can be derived. Detailed optimal solutions are described as follows.
Theorem 2.
For w 1 w 2 w 3 and l 1 + l 2 + l 3 1 , the optimal solution of 3COWAL is as follows:
( y 1 * , y 2 * , y 3 * ) = { ( 1 l 2 l 3 , l 2 , l 3 ) ,   if   l 1 l 2 l 3 ( 1 l 2 l 3 , l 3 , l 2 ) ,   if   l 1 l 3 l 2 ( 1 l 1 l 3 , l 1 , l 3 ) ,   if   l 2 l 1 l 3 ( 1 l 1 l 3 , l 3 , l 1 ) ,   if   l 2 l 3 l 1 ( 1 l 1 l 2 , l 1 , l 2 ) ,   if   l 3 l 1 l 2 ( 1 l 1 l 2 , l 2 , l 1 ) ,   if   l 3 l 2 l 1 .

4. Maximizing Three-Dimensional Constrained OWA Aggregation Problem with Lower Bounded Variables for l1l2l3

This section considers the optimal solution behaviors for 3COWAL with l 1 l 2 l 3 . The main result is described as follows.
Theorem 3.
For l 1 l 2 l 3 and l 1 + l 2 + l 3 1 , the optimal solution Y * of 3COWAL is as follows.
(1) 
For w 1 w 2 w 3 or w 1 w 3 w 2 , the optimal solution is A 1 ( 1 l 2 l 3 , l 2 , l 3 ) .
(2) 
For w 2 w 1 w 3 , the optimal solution Y * is
i f   2 l 1 + l 3 1 ,   t h e n   Y * = A 3 ( l 1 , 1 l 1 l 3 , l 3 )   e l s e   Y * = B 1   ( 1 l 3 2 , 1 l 3 2 , l 3 ) .
(3) 
For w 2 w 3 w 1 , the optimal solution Y * is
i f   2 l 1 + l 3 1 ,   t h e n   Y * = A 3 ( l 1 , 1 l 1 l 3 , l 3 ) .
e l s e   i f   w 3 1 / 3   t h e n   Y * = B 1   ( 1 l 3 2 , 1 l 3 2 , l 3 ) ;
e l s e   i f   l 1 1 / 3 ,   t h e n   Y * = B 5 ( l 1 , l 1 , 1 2 l 1 )   e l s e   Y * = C ( 1 / 3 , 1 / 3 , 1 / 3 ) .
(4) 
For w 3 w 1 w 2 , the optimal solution Y * is
i f   l 1 + 2 l 2 1 ,   t h e n   Y * = A 6 ( l 1 , l 2 , 1 l 1 l 2 ) .
e l s e   i f   w 1 1 / 3   t h e n   Y * = B 4   ( 1 2 l 2 , l 2 , l 2 ) ;
e l s e   i f   l 1 1 / 3   t h e n   Y * = B 3 ( l 1 , 1 l 1 2 , 1 l 1 2 )   e l s e   Y * = C ( 1 / 3 , 1 / 3 , 1 / 3 ) .
(5) 
For w 3 w 2 w 1 , the optimal solution Y * is
i f   l 1 + 2 l 2 1 ,   t h e n   Y * = A 6 ( l 1 , l 2 , 1 l 1 l 2 ) .
e l s e   i f   l 1 1 / 3 ,   t h e n   Y * = B 3 ( l 1 , 1 l 1 2 , 1 l 1 2 )   e l s e   Y * = C ( 1 / 3 , 1 / 3 , 1 / 3 ) .
Proof. 
There are six permutations of ( w 1 , w 2 , w 3 ) . From Theorem 2, for w 1 w 2 w 3 , the optimal solution is A 1 ( 1 l 2 l 3 , l 2 , l 3 ) . Similarly, the optimal solution also is A1 for w 1 w 3 w 2 .
We now consider the case that w 2 w 1 w 3 . From Table 2, all of the twelve candidates are divided into two categories to obtain optimal solution: (I) A1, A3, A4, A5, A6 and C; (II) A2, B1, B2, B3, B4 and B5. We will show that the largest objective function value is that of A 3   ( l 1 , 1 l 1 l 3 , l 3 ) for category I.
Comparing the objective function value of A3 with that of A 1 , from Theorem 1, since w 2 w 1 w 3 , we have to compare s 1 = y 2 , s 2 = y 1 + y 2 and s 3 = y 1 + y 2 + y 3 with those of s 1 = y 2 , s 2 = y 1 + y 2 and s 3 = y 1 + y 2 + y 3 . From
s 1 = 1 l 1 l 3 s 1 = l 2 ,
s 2 = 1 l 3 s 2 = 1 l 3 ,
s 3 = 1 s 3 = 1 ,
the comparison results imply that A3 is superior to A1. Similarly, F ( A 3 ) is larger than those of A2, A5, A6 and C. Therefore, the optimal solution for category I is A3. A similar argument shows that the optimal solution for category II is B 1   ( 1 l 3 2 , 1 l 3 2 , l 3 ) . From the conditions of A3 and B1 displayed in Table 2, the optimal solution for w 2 w 1 w 3 is A3 for 2 l 1 + l 3 1 , and B1 for 2 l 1 + l 3 1 .
Similarly, for w 2 w 3 w 1 , w 3 w 1 w 2 and w 3 w 2 w 1 , the optimal solutions can be derived. □
We next present two numerical experiments to evaluate the optimal solutions of 3COWAL with l 1 l 2 l 3 .
Table 3 and Table 4 entries correspond to a pair (S, W) and give the number of different instances of ( l 1 , l 2 ,   l 3 , w 1 , w 2 ,   w 3 ) satisfying type of candidate solution (S) and weight (W). The adopted measure is the number of instances. From Table 2, we adopt twelve types of candidate solutions and six permutations of weight. More precisely, S { A 1 ,   A 2 ,   A 3 ,   A 4 ,   A 5 ,   A 6 ,   B 1 , B 2 ,   B 3 ,   B 4 ,   B 5 ,   C   } , W { w 1 > w 2 > w 3 ,   w 1 > w 3 > w 2 , w 2 > w 1 > w 3 ,   w 2 > w 3 > w 1 ,   w 3 > w 1 > w 2 ,   w 3 > w 2 > w 1 } , l i { l b , 0.9 , 0.8 , , l b } and w i { 0 , 0.1 , 0.2 , , 1 } , i   =   1 ,   2 ,   3 .
The value of bound l b is l b = 1 for Table 3 and l b = 2 Table 4. For each weight, the instances ( l 1 , l 2 ,   l 3 , w 1 , w 2 ,   w 3 ) of 3COWAL are 8744 for l b = 1 and 57,464 for l b = 2 . The total instances of 3COWAL are 397,248. An examination of Table 3 and Table 4 reveals that the largest number of instances is A 1 for w 1 > w 2 > w 3 and w 1 > w 3 > w 2 , B 1 for w 2 > w 1 > w 3 and w 2 > w 3 > w 1 , and B 3 for w 3 > w 1 > w 2 and w 3 > w 2 > w 1 . Among all the instances of 3COWAL, the zero number is A2, A4, A5 and B2. Therefore, A1, B1 and B3 are superior in the number of the instances, while A2, A4, A5 and B2 are inferior ones for 3COWAL with l 1 l 2 l 3 .

5. Minimizing Three-Dimensional Constrained OWA Aggregation Problem with Upper Bounded Variables

Consider the minimizing COWAU problem described as follows:
Min   W T Y ,   s . t .   I T X 1 ,   G Y 0 ,   y i I X M Z i 0 ,   i = 1 , 2 , , n 1 ,   y n I X 0 ,   I T Z i n i ,   i = 1 , 2 , , n 1 ,   Z i + 1 Z i ,   i = 1 , 2 , , n 2 ,   Z i { 0 , 1 } n ,   i = 1 , 2 , , n 1 ,   X U
where the column vector
U = [ u 1 u 2 u n ] .
Using the standard transformations X = U X , these lead to the following model
Min   W T Y ,   s . t .   I T X I T U 1 ,   G Y 0 ,   y i I + X M Z i U ,   i = 1 , 2 , , n 1 ,   y n I + X U ,   I T Z i n i ,   i = 1 , 2 , , n 1 ,   Z i + 1 Z i ,   i = 1 , 2 , , n 2 ,   Z i { 0 , 1 } n ,   i = 1 , 2 , , n 1 ,   X 0 .
If I T U 1 < 0 , we conclude that the COWAU has no feasible solutions. If I T U 1 = 0 , then X = 0 is the unique optimal solution, so X = U .
This section considers 3COWAU for
I T U 1 = u 1 + u 2 + u 3 1 .
Similar analyses to Section 3 and Section 4 can be derived. The results are described as follows.
At optimality, the first constraint of the model (13) becomes
  x 1 + x 2 + x 3 = u 1 + u 2 + u 3 1 .  
There are three types (A′, B′, C′) of ( x 1 , x 2 , x 3 ) according to the number of zero components.
For 3COWAU with
u 1 u 2 u 3 ,
there are six candidate optimal solutions for type A′, seven candidate optimal solutions for type B′ and one candidate optimal solution for type C′. Detailed results of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) , ( y 1 , y 2 , y 3 ) , F ( y 1 , y 2 , y 3 ) and condition are presented in Table 5.
Theorem 4.
For 3COWAU with u 1 u 2 u 3 and w 1 w 2 w 3 , the smallest objective function value is that of ( y 1 , y 2 , y 3 ) = A 1 ( 1 u 2 u 3 , u 2 , u 3 ) for type A′, B 1 ( 1 u 3 2 , 1 u 3 2 , u 3 ) for type B′ and C ( 1 / 3 , 1 / 3 , 1 / 3 ) for type C′.
For 3COWAU with u 1 u 2 u 3 , the optimal solutions of different permutations of weight are described as follows.
Theorem 5.
For 3COWAU with u 1 u 2 u 3 and u 1 + u 2 + u 3 1 , the optimal solution Y * is described as follows.
(1) 
For w 1 w 2 w 3 , the optimal solution Y * is
i f   u 3 1 / 3 ,   t h e n   Y * = C ( 1 3 , 1 3 , 1 3 )   e l s e   i f   2 u 2 + u 3 1 t h e n   Y * = A 1 ( 1 u 2 u 3 , u 2 , u 3 )   e l s e   Y * = B 1 ( 1 u 3 2 , 1 u 3 2 , u 3 ) .
(2) 
For w 1 w 3 w 2 , the optimal solution Y * is
i f   2 u 2 + u 3 1 ,   t h e n   Y * = A 1   ( 1 u 2 u 3 , u 2 , u 3 ) .
e l s e   i f   w 3 1 / 3 ,   t h e n   Y * = B 7   ( u 2 , u 2 , 1 2 u 2 ) ;
e l s e   i f   u 3 1 / 3 ,   t h e n   Y * = B 1 ( 1 u 3 2 , 1 u 3 2 , u 3 )   e l s e   Y * = C ( 1 3 , 1 3 , 1 3 ) .
(3) 
For w 2 w 1 w 3 , the optimal solution Y * is
i f   u 1 + 2 u 3 1 ,   t h e n   Y * = A 4   ( u 1 , 1 u 1 u 3 , u 3 ) .
e l s e   i f   w 1 1 / 3 ,   t h e n   Y * = B 4   ( u 1 , 1 u 1 2 , 1 u 1 2 ) ;
e l s e   i f   u 3 1 / 3 ,   t h e n   Y * = B 6 ( 1 2 u 3 , u 3 , u 3 )   e l s e   Y * = C ( 1 3 , 1 3 , 1 3 ) .
(4) 
For w 2 w 3 w 1 , the optimal solution Y * is
i f   u 1 + 2 u 3 1 ,   t h e n   Y * = A 4 ( u 1 , 1 u 1 u 3 , u 3 )   e l s e   Y * = B 4 ( u 1 , 1 u 1 2 , 1 u 1 2 ) .
(5) 
For w 3 w 1 w 2 and w 3 w 2 w 1 , the optimal solution is A 6 ( u 1 , u 2 , 1 u 1 u 2 ) .
To compare the optimal solution behaviors of maximizing 3COWAL with those of minimizing 3COWAU, the performance is shown in Figure 1. The first bar corresponds to the number of optimal solutions of maximizing 3COWAL and second to the minimizing 3COWAU, where the weight type W [ i ] denotes the ith component of W = { w 1 > w 2 > w 3 ,   w 1 > w 3 > w 2 , w 2 > w 1 > w 3 ,   w 2 > w 3 > w 1 ,   w 3 > w 1 > w 2 ,   w 3 > w 2 > w 1 } . The number of optimal solutions of W [ i ] , i = 1 ,   2 ,   , 6 , for maximizing 3COWAL is the same as that of W [ 6   i ] for minimizing 3COWAU. Therefore, the numbers of optimal solutions for maximizing 3COWAL are same as those of minimizing 3COWAU but in reverse order. The correspondence between the optimal solution of these two mathematical models is worthy of future research.
We perform two numerical experiments to evaluate the optimal solution behaviors of 3COWAU with u 1 u 2 u 3 .
Table 6 and Table 7 give the number of different instances ( u 1 , u 2 ,   u 3 , w 1 , w 2 ,   w 3 ) satisfying type of candidate solution (S) and weight (W). More precisely,
S { A 1 ,   A 2 ,   ,   A 6 ,   B 1 , B 2 ,   ,   B 7 ,   C } ,   W { w 1 > w 2 > w 3 ,   w 1 > w 3 > w 2 ,   w 2 > w 1 > w 3 ,   w 2 > w 3 > w 1 ,   w 3 > w 1 > w 2 ,   w 3 > w 2 > w 1 } ,   u i { u b , 0.9 , 0.8 , , u b }   and   w i { 0 ,   0.1 ,   0.2 , ,   1 } i = 1 ,   2 ,   3 .
The value of bound u b is u b = 1 for Table 6 and u b = 2 Table 7. For each weight, the number of the instances of 3COWAU is 8744 for u b = 1 and 57,464 for u b = 2 . The total instances of 3COWAU are 397,248. From Table 6 and Table 7, the largest number of instances is B 1 ( 1 u 3 2 , 1 u 3 2 , u 3 ) for w 1 > w 2 > w 3 , A 1   ( 1 u 2 u 3 , u 2 , u 3 ) for w 1 > w 3 > w 2 , A 4   ( u 1 , 1 u 1 u 3 , u 3 ) for w 2 > w 1 > w 3 and w 2 > w 3 > w 1 , and A 6 ( u 1 , u 2 , 1 u 1 u 2 ) for w 3 > w 1 > w 2 and w 3 > w 2 > w 1 . We also notice that the number of instances is zero for A′2, A′3, A′5, B′2, B′3 and B′5. Therefore, A 1 , A 4 , A 6 and B 1 are the best candidates, while A′2, A′3, A′5, B′2, B′3 and B′5 are inferior ones for 3COWAU with u 1 u 2 u 3 .

6. Conclusions

This paper presents the optimal solutions for both maximizing 3COWAL and minimizing 3COWAU. For maximizing 3COWAL with l 1 l 2 l 3 , there are six candidate optimal solutions for type A, five candidate optimal solutions for type B and one candidate optimal solution for type C. Theoretically and empirically, the largest number of instances is A 1   ( 1 l 2 l 3 , l 2 , l 3 ) for w 1 > w 2 > w 3 and w 1 > w 3 > w 2 , B 1   ( 1 l 3 2 , 1 l 3 2 , l 3 ) for w 2 > w 1 > w 3 and w 2 > w 3 > w 1 , and B 3   ( l 1 , 1 l 1 2 , 1 l 1 2 ) for w 3 > w 1 > w 2 and w 3 > w 2 > w 1 . For minimizing 3COWAU with u 1 u 2 u 3 , there are six candidate optimal solutions for type A′, seven candidate optimal solutions for type B′ and one candidate optimal solution for type C′. The largest number of instances is B 1   ( 1 u 3 2 , 1 u 3 2 , u 3 ) for w 1 > w 2 > w 3 , A 1   ( 1 u 2 u 3 , u 2 , u 3 ) for w 1 > w 3 > w 2 , A 4   ( u 1 , 1 u 1 u 3 , u 3 ) for w 2 > w 1 > w 3 and w 2 > w 3 > w 1 , and A 6   ( u 1 , u 2 , 1 u 1 u 2 ) for w 3 > w 1 > w 2 and w 3 > w 2 > w 1 . Therefore, the best candidate optimal solutions are A1, B1 and B3 for maximizing 3COWAL with l 1 l 2 l 3 , and A 1 , A 4 , A 6 and B 1 for minimizing 3COWAU with u 1 u 2 u 3 .
Extending the analysis to high dimensions is worthy of future research in addition to analysis of the correspondence between the optimal solution of maximizing 3COWAL and minimizing 3COWAU. Thus, the analysis of maximizing COWAL and minimizing COWAU is a subject of considerable ongoing research.

Author Contributions

H.T. analyzed the method and wrote the paper. S.Y. performed the experiments.

Funding

This research received no external funding.

Conflicts of Interest

The authors declare that they have no competing interests.

References

  1. Yager, R.R. On ordered weighted averaging aggregation operators in multicriteria decision making. IEEE Trans. Syst. Man Cybern. 1988, 18, 183–190. [Google Scholar] [CrossRef]
  2. Chaji, A.; Fukuyama, H.; Shiraz, R.K. Selecting a model for generating OWA operator weights in MAGDM problems by maximum entropy membership function. Comp. Ind. Eng. 2018, 124, 370–378. [Google Scholar] [CrossRef]
  3. De Miguel, L.; Sesma-Sara, M.; Elkano, M.; Asiain, M.; Bustince, H. An algorithm for group decision making using n-dimensional fuzzy sets, admissible orders and OWA operators. Inf. Fusion 2017, 37, 126–131. [Google Scholar] [CrossRef]
  4. Liu, P.; Wang, Y. Interval neutrosophic prioritized OWA operator and its application to multiple attribute decision making. J. Syst. Sci. Complex. 2016, 29, 681–697. [Google Scholar] [CrossRef]
  5. Merigó, J.M.; Gil-Lafuente, A.M.; Zhou, L.G.; Chen, H.Y. Induced and linguistic generalized aggregation operators and their application in linguistic group decision making. Group Decis. Negot. 2012, 21, 531–549. [Google Scholar] [CrossRef]
  6. Zhou, L.; Chen, H.; Liu, J. Generalized multiple averaging operators and their applications to group decision making. Group Decis. Negot. 2013, 22, 331–358. [Google Scholar] [CrossRef]
  7. Vigier, H.P.; Scherger, V.; Terceño, A. An application of OWA operators in fuzzy business diagnosis. Appl. Soft Comput. 2017, 54, 440–448. [Google Scholar] [CrossRef]
  8. Yager, R.R. On the analytic representation of the Leximin ordering and its application to flexible constraint propagation. Eur. J. Oper. Res. 1997, 102, 176–192. [Google Scholar] [CrossRef]
  9. Garg, B.; Garg, R. Enhanced accuracy of fuzzy time series model using ordered weighted aggregation. Appl. Soft Comput. 2016, 48, 265–280. [Google Scholar] [CrossRef]
  10. Mohammed, E.A.; Naugler, C.T.; Far, B.H. Breast tumor classification using a new OWA operator. Exp. Syst. Appl. 2016, 61, 302–313. [Google Scholar] [CrossRef]
  11. Yager, R.R.; Beliakov, G. OWA operators in regression problems. IEEE Trans. Fuzzy Syst. 2010, 18, 106–113. [Google Scholar] [CrossRef]
  12. Ogryczak, W.; Śliwiński, T. On solving linear programs with the ordered weighted averaging objective. Eur. J. Oper. Res. 2003, 148, 80–91. [Google Scholar] [CrossRef]
  13. Ameri, F.; Zoej, M.J.V.; Mokhtarzade, M. Multi-Criteria, Graph-Based Road Centerline Vectorization Using Ordered Weighted Averaging Operators. Photogramm. Eng. Remote Sens. 2016, 82, 107–120. [Google Scholar] [CrossRef]
  14. Chen, S.M.; Kuo, L.W. Autocratic decision making using group recommendations based on interval type-2 fuzzy sets, enhanced Karnik–Mendel algorithms, and the ordered weighted aggregation operator. Inf. Sci. 2017, 412, 174–193. [Google Scholar] [CrossRef]
  15. Carlsson, C.; Fullér, R. Maximal Entropy and Minimal Variability OWA Operator Weights: A Short Survey of Recent Developments. In Soft Computing Applications for Group Decision-Making and Consensus Modeling; Mikael, C., Janusz, K., Eds.; Springer: Cham, Switzerland, 2018; pp. 187–199. [Google Scholar]
  16. Emrouznejad, A.; Marra, M. Ordered weighted averaging operators 1988–2014: A citation-based literature survey. Int. J. Intell. Syst. 2014, 29, 994–1014. [Google Scholar] [CrossRef]
  17. Yager, R.R.; Kacprzyk, J.; Beliakov, G. Recent Developments in the Ordered Weighted Averaging Operators: Theory and Practice; Springer: Berlin, Germany, 2011. [Google Scholar]
  18. Yager, R.R. Constrained OWA aggregation. Fuzzy Sets Syst. 1996, 81, 89–101. [Google Scholar] [CrossRef]
  19. Carlsson, C.; Fullér, R.; Majlender, P. A note on constrained OWA aggregation. Fuzzy Sets Syst. 2003, 139, 543–546. [Google Scholar] [CrossRef]
  20. Coroianu, L.; Fullér, R. On the constrained OWA aggregation problem with single constraint. Fuzzy Sets Syst. 2018, 332, 37–43. [Google Scholar] [CrossRef]
  21. Ahn, B.S. Parameterized OWA operator weights: An extreme point approach. Int. J. Approx. Reason. 2010, 51, 820–831. [Google Scholar] [CrossRef]
  22. Ahn, B.S. A New Approach to Solve the Constrained OWA Aggregation Problem. IEEE Trans. Fuzzy Syst. 2017, 25, 1231–1238. [Google Scholar] [CrossRef]
  23. Fernández, E.; Pozo, M.A.; Puerto, J.; Scozzari, A. Ordered Weighted Average optimization in Multiobjective Spanning Tree Problem. Eur. J. Oper. Res. 2017, 260, 886–903. [Google Scholar] [CrossRef]
  24. Fullér, R.; Majlender, P. On obtaining minimal variability OWA operator weights. Fuzzy Sets Syst. 2003, 136, 203–215. [Google Scholar] [CrossRef]
  25. Kasperski, A.; Zieliński, P. Using the WOWA operator in robust discrete optimization problems. Int. J. Approx. Reason. 2016, 68, 54–67. [Google Scholar] [CrossRef]
  26. Ogryczak, W.; Olender, P. Ordered median problem with demand distribution weights. Optim. Lett. 2016, 10, 1071–1086. [Google Scholar] [CrossRef]
  27. Ogryczak, W.; Śliwiński, T. On efficient WOWA optimization for decision support under risk. Int. J. Approx. Reason. 2009, 50, 915–928. [Google Scholar] [CrossRef]
  28. Chen, Y.F.; Tang, H.C. A Three Dimensional Constrained OWA Aggregation Problem with Lower Bounded Variables. Symmetry 2018, 10, 339. [Google Scholar] [CrossRef]
Figure 1. The number of optimal solutions for maximizing 3COWAL and minimizing 3COWAU.
Figure 1. The number of optimal solutions for maximizing 3COWAL and minimizing 3COWAU.
Mathematics 06 00172 g001
Table 1. The possible values of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) and ( y 1 , y 2 , y 3 ) for 3COWAL with l 1 l 2 l 3 .
Table 1. The possible values of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) and ( y 1 , y 2 , y 3 ) for 3COWAL with l 1 l 2 l 3 .
( x 1 , x 2 , x 3 ) ( x 1 , x 2 , x 3 ) ( y 1 , y 2 , y 3 )
A-1 ( 1 l 1 l 2 l 3 , 0 ,   0 ) ( 1 l 2 l 3 ,   l 2 ,   l 3 ) ( 1 l 2 l 3 ,   l 2 ,   l 3 )
A-2 ( 0 ,   1 l 1 l 2 l 3 , 0 ) ( l 1 , 1 l 1 l 3 ,   l 3 ) ( 1 l 1 l 3 ,   l 1 ,   l 3 ) , ( l 1 , 1 l 1 l 3 ,   l 3 )
A-3 ( 0 ,   0 ,   1 l 1 l 2 l 3 ) ( l 1 ,   l 2 ,   1 l 1 l 2 ) ( 1 l 1 l 2 ,   l 1 ,   l 2 ) , ( l 1 ,   1 l 1 l 2 ,   l 2 ) ,   ( l 1 ,   l 2 ,   1 l 1 l 2 )
B1-1 ( 1 2 l 1 l 3 2 ,   1 2 l 2 l 3 2 ,   0 ) ( 1 l 3 2 ,   1 l 3 2 ,   l 3 ) ( 1 l 3 2 , 1 l 3 2 , l 3 )
B1-2 ( 1 2 l 1 l 2 2 , 0 , 1 l 2 2 l 3 2 ) ( 1 l 2 2 , l 2 , 1 l 2 2 ) ( 1 l 2 2 , 1 l 2 2 , l 2 )
B1-3 ( 0 , 1 l 1 2 l 2 2 , 1 l 1 2 l 3 2 ) ( l 1 , 1 l 1 2 , 1 l 1 2 ) ( l 1 , 1 l 1 2 , 1 l 1 2 )
B2-1 ( l 3 l 1 , 1 l 2 2 l 3 , 0 ) ( l 3 , 1 2 l 3 , l 3 )
B2-2 ( 1 l 1 2 l 3 , l 3 l 2 , 0 ) ( 1 2 l 3 , l 3 ,   l 3 )
B2-3 ( l 2 l 1 , 0 ,   1 2 l 2 l 3 ) ( l 2 , l 2 , 1 2 l 2 )
B2-4 ( 1 l 1 2 l 2 , 0 ,   l 2 l 3 ) ( 1 2 l 2 , l 2 ,   l 2 ) ( 1 2 l 2 ,   l 2 ,   l 2 )
B2-5 ( 0 ,   l 1 l 2 , 1 2 l 1 l 3 ) ( l 1 , l 1 , 1 2 l 1 ) ( l 1 ,   l 1 , 1 2 l 1 )
B2-6 ( 0 ,   1 2 l 1 l 2 ,   l 1 l 3 ) ( l 1 , 1 2 l 1 , l 1 ) ( l 1 ,   l 1 , 1 2 l 1 )
C ( 1 / 3 l 1 , 1 / 3 l 2 , 1 / 3 l 3 ) ( 1 / 3 ,   1 / 3 ,   1 / 3 ) ( 1 / 3 ,   1 / 3 ,   1 / 3 )
Table 2. The candidate optimal solutions of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) and ( y 1 , y 2 , y 3 ) , F ( y 1 , y 2 , y 3 ) and condition for 3COWAL with l 1 l 2 l 3 .
Table 2. The candidate optimal solutions of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) and ( y 1 , y 2 , y 3 ) , F ( y 1 , y 2 , y 3 ) and condition for 3COWAL with l 1 l 2 l 3 .
  ( x 1 , x 2 , x 3 )     ( x 1 , x 2 , x 3 ) ( y 1 , y 2 , y 3 ) F ( y 1 , y 2 , y 3 ) Condition
A1 ( 1 l 1 l 2 l 3 , 0 ,   0 ) ( 1 l 2 l 3 ,   l 2 ,   l 3 ) ( 1 l 2 l 3 ,   l 2 ,   l 3 ) w 1 + l 2 (   w 1 + w 2 ) + l 3 ( w 1 + w 3 )
A2 ( 0 , 1 l 1 l 2 l 3 , 0 ) ( l 1 , 1 l 1 l 3 ,   l 3 ) ( 1 l 1 l 3 ,   l 1 ,   l 3 ) w 1 + l 1 ( w 1 + w 2 ) + l 3 ( w 1 + w 3 ) 2 l 1 + l 3 1
A3 ( 0 , 1 l 1 l 2 l 3 , 0 ) ( l 1 , 1 l 1 l 3 ,   l 3 ) ( l 1 , 1 l 1 l 3 ,   l 3 ) w 2 + l 1 ( w 1 w 2 ) + l 3 (   w 2 + w 3 ) 2 l 1 + l 3 1
A4 ( 0 ,   0 ,   1 l 1 l 2 l 3 ) ( l 1 ,   l 2 , 1 l 1 l 2 ) ( 1 l 1 l 2 ,   l 1 ,   l 2 ) w 1 + l 1 (   w 1 + w 2 ) + l 2 (   w 1 + w 3 ) 2 l 1 + l 2 1
A5 ( 0 ,   0 ,   1 l 1 l 2 l 3 ) ( l 1 ,   l 2 , 1 l 1 l 2 ) ( l 1 , 1 l 1 l 2 ,   l 2 ) w 2 + l 1 ( w 1 w 2 ) + l 2 ( w 2 + w 3 ) 2 l 1 + l 2 1 , l 1 + 2 l 2 1
A6 ( 0 ,   0 ,   1 l 1 l 2 l 3 ) ( l 1 ,   l 2 , 1 l 1 l 2 ) ( l 1 ,   l 2 , 1 l 1 l 2 ) w 3 + l 1 ( w 1 w 3 ) + l 2 ( w 2 w 3 ) l 1 + 2 l 2 1
B1 ( 1 2 l 1 l 3 2 , 1 2 l 2 l 3 2 , 0 ) ( 1 l 3 2 , 1 l 3 2 ,   l 3 ) ( 1 l 3 2 , 1 l 3 2 ,   l 3 ) 1 w 3 l 3 + 3 l 3 w 3 2 l 3 1 / 3 , 2 l 1 + l 3 1
B2 ( 1 2 l 1 l 2 2 , 0 , 1 l 2 2 l 3 2 ) ( 1 l 2 2 ,   l 2 , 1 l 2 2 ) ( 1 l 2 2 , 1 l 2 2 ,   l 2 ) 1 w 3 l 2 + 3 l 2 w 3 2 l 2 1 / 3 , 2 l 1 + l 2 1
B3 ( 0 , 1 l 1 2 l 2 2 , 1 l 1 2 l 3 2 ) ( l 1 , 1 l 1 2 , 1 l 1 2 ) ( l 1 , 1 l 1 2 , 1 l 1 2 ) 1 w 1 l 1 + 3 l 1 w 1 2 l 1 1 / 3 , l 1 + 2 l 2 1
B4 ( 1 l 1 2 l 2 , 0 , l 2 l 3 ) ( 1 2 l 2 ,   l 2 ,   l 2 )   ( 1 2 l 2 ,   l 2 ,   l 2 ) w 1 + l 2 3 l 2 w 1 l 2 1 / 3 , l 1 + 2 l 2 1
B5 ( 0 ,   l 1 l 2 ,   1 2 l 1 l 3 ) ( l 1 ,   l 1 , 1 2 l 1 ) ( l 1 ,   l 1 , 1 2 l 1 ) w 3 + l 1 3 l 1 w 3 l 1 1 / 3 , 2 l 1 + l 3 1
C ( 1 / 3 l 1 ,   1 / 3 l 2 ,   1 / 3 l 3 ) ( 1 / 3 ,   1 / 3 ,   1 / 3 ) ( 1 / 3 ,   1 / 3 ,   1 / 3 ) 1/3 l 1 1 / 3 , l 2 1 / 3 , l 3 1 / 3
Table 3. The number of instances satisfying solution type and weight for 3COWAL with 1 l 3 < l 2 < l 1 1 .
Table 3. The number of instances satisfying solution type and weight for 3COWAL with 1 l 3 < l 2 < l 1 1 .
( y 1 , y 2 , y 3 ) w 1 > w 2 > w 3 w 1 > w 3 > w 2 w 2 > w 1 > w 3 w 2 > w 3 > w 1 w 3 > w 1 > w 2 w 3 > w 2 > w 1 Total
A1 ( 1 l 2 l 3 ,   l 2 ,   l 3 ) 87448744000017,488
A2 ( 1 l 1 l 3 ,   l 1 ,   l 3 ) 0000000
A3 ( l 1 , 1 l 1 l 3 ,   l 3 ) 0016321632003264
A4 ( 1 l 1 l 2 ,   l 1 ,   l 2 ) 0000000
A5 ( l 1 , 1 l 1 l 2 ,   l 2 ) 0000000
A6 ( l 1 ,   l 2 , 1 l 1   l 2 ) 0000192019203840
B1 ( 1 l 3 2 , 1 l 3 2 ,   l 3 ) 00711254740012,586
B2 ( 1 l 2 2 , 1 l 2 2 ,   l 2 ) 0000000
B3 ( l 1 , 1 l 1 2 , 1 l 1 2 ) 0000302639126938
B4   ( 1 2 l 2 ,   l 2 ,   l 2 ) 0000161401614
B5 ( l 1 ,   l 1 , 1 2 l 1 ) 00091000910
C ( 1 / 3 ,   1 / 3 ,   1 / 3 ) 000728218429125824
Table 4. The number of instances satisfying solution type and weight for 3COWAL with 2 l 3 < l 2 < l 1 2 .
Table 4. The number of instances satisfying solution type and weight for 3COWAL with 2 l 3 < l 2 < l 1 2 .
( y 1 , y 2 , y 3 ) w 1 > w 2 > w 3 w 1 > w 3 > w 2 w 2 > w 1 > w 3 w 2 > w 3 > w 1 w 3 > w 1 > w 2 w 3 > w 2 > w 1 Total
A1 ( 1 l 2 l 3 ,   l 2 ,   l 3 ) 57,46457,4640000114,928
A2 ( 1 l 1 l 3 ,   l 1 ,   l 3 ) 0000000
A3 ( l 1 , 1 l 1 l 3 ,   l 3 ) 0019,35219,3520038,704
A4 ( 1 l 1 l 2 ,     l 1 ,   l 2 ) 0000000
A5 ( l 1 , 1 l 1 l 2 ,   l 2 ) 0000000
A6 ( l 1 ,   l 2 , 1 l 1 l 2 ) 000015,64017,00832,648
B1 ( 1 l 3 2 , 1 l 3 2 ,   l 3 ) 0038,11229,0040067,116
B2 ( 1 l 2 2 , 1 l 2 2 ,   l 2 ) 0000000
B3 ( l 1 ,   1 l 1 2 , 1 l 1 2 ) 000019,56628,31247,878
B4   ( 1 2 l 2 ,   l 2 ,   l 2 ) 000010,114010,114
B5 ( l 1 ,   l 1 , 1 2 l 1 ) 0005060005060
C ( 1 / 3 ,   1 / 3 ,   1 / 3 ) 000404812,14412,14428,336
Table 5. The candidate optimal solutions of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) and ( y 1 , y 2 , y 3 ) , F ( y 1 , y 2 , y 3 ) and condition for 3COWAU with u 1 u 2 u 3 .
Table 5. The candidate optimal solutions of ( x 1 , x 2 , x 3 ) , ( x 1 , x 2 , x 3 ) and ( y 1 , y 2 , y 3 ) , F ( y 1 , y 2 , y 3 ) and condition for 3COWAU with u 1 u 2 u 3 .
( x 1 , x 2 , x 3 ) ( x 1 , x 2 , x 3 ) ( y 1 , y 2 , y 3 ) F ( y 1 , y 2 , y 3 ) Condition
A′1 ( u 1 + u 2 + u 3 1 ,   0 ,   0 ) ( 1 u 2 u 3 ,   u 2 ,   u 3 ) ( 1 u 2 u 3 ,   u 2 ,   u 3 ) w 1 + u 2 ( w 1 + w 2 ) + u 3 ( w 1 + w 3 ) 2 u 2 + u 3 1
A′2 ( u 1 + u 2 + u 3 1 ,   0 ,   0 ) ( 1 u 2 u 3 ,   u 2 ,   u 3 ) ( u 2 , 1 u 2 u 3 , u 3 ) w 2 + u 2 ( w 1 w 2 ) + u 3 ( w 2 + w 3 ) 2 u 2 + u 3 1 , u 2 + 2 u 3 1
A′3 ( u 1 + u 2 + u 3 1 ,   0 ,   0 ) ( 1 u 2 u 3 ,   u 2 ,   u 3 ) ( u 2 , u 3 , 1 u 2 u 3 ) w 3 + u 2 ( w 1 w 3 ) + u 3 ( w 2 w 3 ) u 2 + 2 u 3 1 u 2 1 / 3
A′4 ( 0 ,   u 1 + u 2 + u 3 1 ,   0 ) ( u 1 , 1 u 1 u 3 ,   u 3 ) ( u 1 , 1 u 1 u 3 , u 3 ) w 2 + u 1 ( w 1 w 2 ) + u 3 ( w 2 + w 3 ) u 1 + 2 u 3 1 , u 3 1 / 3
A′5 ( 0 ,   u 1 + u 2 + u 3 1 ,   0 ) ( u 1 , 1 u 1 u 3 ,   u 3 ) ( u 1 ,   u 3 , 1 u 1 u 3 ) w 3 + u 1 ( w 1 w 3 ) + u 3 ( w 2 w 3 ) u 1 + 2 u 3 1
A′6 ( 0 ,   0 ,   u 1 + u 2 + u 3 1 ) ( u 1 , u 2 , 1 u 1 u 2 ) ( u 1 ,   u 2 , 1 u 1 u 2 ) w 3 + u 1 ( w 1 w 3 ) + u 2 ( w 2 w 3 )
B′1 ( 2 u 1 + u 3 1 2 ,   2 u 2 + u 3 1 2 , 0 ) ( 1 u 3 2 , 1 u 3 2 ,   u 3 ) ( 1 u 3 2 , 1 u 3 2 ,   u 3 ) 1 w 3 u 3 + 3 u 3 w 3 2 u 3 1 / 3 , 2 u 2 + u 3 1
B′2 ( 2 u 1 + u 3 1 2 , 2 u 2 + u 3 1 2 , 0 ) ( 1 u 3 2 , 1 u 3 2 ,   u 3 ) ( u 3 , 1 u 3 2 , 1 u 3 2 ) 1 w 1 u 3 + 3 u 3 w 1 2 u 3 1 / 3
B′3 ( 2 u 1 + u 2 1 2 , 0 , 2 u 3 + u 2 1 2 ) ( 1 u 2 2 ,   u 2 , 1 u 2 2 ) ( u 2 , 1 u 2 2 , 1 u 2 2 ) 1 w 1 u 2 + 3 u 2 w 1 2 u 2 1 / 3 , u 2 + 2 u 3 1
B′4 ( 0 , 2 u 2 + u 1 1 2 , 2 u 3 + u 1 1 2 ) ( u 1 , 1 u 1 2 , 1 u 1 2 ) ( u 1 , 1 u 1 2 , 1 u 1 2 ) 1 w 1 u 1 + 3 u 1 w 1 2 u 1 + 2 u 3 1
B′5 ( u 1 u 3 , u 2 + 2 u 3 1 ,   0 ) ( u 3 , 1 2 u 3 ,   u 3 ) ( u 3 ,   u 3 , 1 2 u 3 ) w 3 + u 3 3 u 3 w 3 u 3 1 / 3
B′6 ( u 1 u 3 , u 2 + 2 u 3 1 ,   0 ) ( u 3 , 1 2 u 3 ,   u 3 ) ( 1 2 u 3 ,   u 3 ,   u 3 ) w 1 + u 3 3 u 3 w 1 u 3 1 / 3 , u 2 + 2 u 3 1
B′7 ( u 1 u 2 , 0 , 2 u 2 + u 3 1 ) ( u 2 , u 2 ,   1 2 u 2 ) ( u 2 ,   u 2 , 1 2 u 2 ) w 3 + u 2 3 u 2 w 3 u 2 1 / 3 , 2 u 2 + u 3 1
C′ ( u 1 1 / 3 , u 2 1 / 3 , u 3 1 / 3 ) ( 1 / 3 ,   1 / 3 ,   1 / 3 ) ( 1 / 3 ,   1 / 3 ,   1 / 3 ) 1/3 u 1 1 / 3 , u 2 1 / 3 , u 3 1 / 3
Table 6. The number of instances satisfying solution type and weight for 3COWAU with 1   u 3 u 2 u 1 1 .
Table 6. The number of instances satisfying solution type and weight for 3COWAU with 1   u 3 u 2 u 1 1 .
( y 1 , y 2 , y 3 ) w 1 > w 2 > w 3 w 1 > w 3 > w 2 w 2 > w 1 > w 3 w 2 > w 3 > w 1 w 3 > w 1 > w 2 w 3 > w 2 > w 1 Total
A′1 ( 1 u 2 u 3 , u 2 , u 3 ) 62462400001248
A′2 ( u 2 , 1 u 2 u 3 , u 3 ) 0000000
A′3 ( u 2 , u 3 , 1 u 2 u 3 ) 0000000
A′4 ( u 1 , 1 u 1 u 3 , u 3 ) 00912888001800
A′5 ( u 1 ,   u 3 ,   1 u 1 u 3 ) 0000000
A′6 ( u 1 , u 2 ,   1 u 1 u 2 ) 0000163216323264
B′1 ( 1 u 3 2 ,   1 u 3 2 ,   u 3 ) 72854600001274
B′2 ( u 3 ,   1 u 3 2 ,   1 u 3 2 ) 0000000
B′3 ( u 2 ,   1 u 2 2 ,   1 u 2 2 ) 0000000
B′4 ( u 1 ,   1 u 1 2 ,   1 u 1 2 ) 00558744001302
B′5 ( u 3 , u 3 , 1 2 u 3 ) 0000000
B′6 ( 1 2 u 3 ,   u 3 ,   u 3 ) 009200092
B′7 ( u 2 ,   u 2 ,   1 2 u 2 ) 02520000252
C′ ( 1 / 3 ,   1 / 3 ,   1 / 3 ) 28021070000560
Table 7. The number of instances satisfying solution type and weight for 3COWAU with 2   u 3 u 2 u 1 2 .
Table 7. The number of instances satisfying solution type and weight for 3COWAU with 2   u 3 u 2 u 1 2 .
( y 1 , y 2 , y 3 ) w 1 > w 2 > w 3 w 1 > w 3 > w 2 w 2 > w 1 > w 3 w 2 > w 3 > w 1 w 3 > w 1 > w 2 w 3 > w 2 > w 1 Total
A′1 ( 1 u 2 u 3 ,   u 2 ,   u 3 ) 91849184000018,368
A′2 ( u 2 , 1 u 2 u 3 ,   u 3 ) 0000000
A′3 ( u 2 ,   u 3 ,   1 u 2 u 3 ) 0000000
A′4 ( u 1 ,   1 u 1 u 3 ,   u 3 ) 0013,47213,3280026,800
A′5 ( u 1 ,   u 3 ,   1 u 1 u 3 ) 0000000
A′6 ( u 1 ,   u 2 ,   1 u 1 u 2 ) 000026,35226,35252,704
B′1 ( 1 u 3 2 ,   1 u 3 2 ,   u 3 ) 11,7288796000020,524
B′2 ( u 3 , 1 u 3 2 , 1 u 3 2 ) 0000000
B′3 ( u 2 ,   1 u 2 2 ,   1 u 2 2 ) 0000000
B′4 ( u 1 ,   1 u 1 2 ,   1 u 1 2 ) 00976813,0240022,792
B′5 ( u 3 ,   u 3 ,   1 2 u 3 ) 0000000
B′6 ( 1 2 u 3 ,   u 3 , u 3 ) 0017520001752
B′7 ( u 2 ,   u 2 ,   1 2 u 2 ) 0429200004292
C′ ( 1 / 3 ,   1 / 3 ,   1 / 3 ) 54404080136000010,880

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Mathematics EISSN 2227-7390 Published by MDPI AG, Basel, Switzerland RSS E-Mail Table of Contents Alert
Back to Top