International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
139
Faculty Recruitment in Engineering Organization Through Fuzzy
Multi-Criteria Group Decision Making Methods
Ruchika Banerjee
1
and Dipendra Nath Ghosh
2
1
Department of Information Technology
Dr. B. C Roy Engineering College, Durgapur-713206, West Bengal, India
2
Department of Computer Science & Engineering
Dr B. C Roy Engineering College, Durgapur-713206, West Bengal, India
[Link]@[Link], ghoshdipen2003@[Link]
Abstract
Faculty selection procedure in any educational organization is a multi-criteria decision
making (MCDM) problem. Faculty members play an important role while judging the quality
of an educational institute. Thus, educational organizations are interested to recruit the best
faculty in order to provide a quality education to their students. As, this MCDM problem
requires multiple criteria and a finite set of candidate alternatives, an amalgamation of
various MCDM methods can be used to solve such problem. In this paper a study has been
made on ranking five alternatives with the help of three step methodology combining FAHP
with PROMETHEE-2 and TOPSIS, comparing the results and finally determining the rank
using group decision making method. This proposed model yields the ranking of the five
candidates in a faculty interview considering several experts opinion.
Keywords: MCDM, Fuzzy AHP, PROMETHEE-2, TOPSIS, Spearmans Rank Correlation
Coefficient, Group Decision Making
1. Introduction
Education System plays a vital role in the development of any country. As, best
educational institutes are an asset for a country, in the same way best faculty members are an
asset for that institute. The quality of faculty members in an educational organization
determines the quality of education provided to their students. Thus choosing the best faculty
staff members for their institute have become a major priority among the selectors. Similar
study regarding the Faculty selection in Engineering Organization is done in [1] and it forms
the basis of this paper. To improve the lack of recruitment processes as well as reduce
individual senses of supervisory level by fuzzy logic and AHP methods, Pin-Chang Chen in
[2] tried to identify appropriate personality traits and key professional skills through the
information statistics. Kazem Oraee and et al., combined AHP with TOPSIS and
PROMETHEE-2 in two step methodology for selecting a tunnel system [3]. Evaluation of
best technical institutions fuzzy analytic hierarchy process was developed to tolerate
vagueness and uncertainty of human judgment [4].
This paper efficiently combines fuzzy analytic hierarchy process with PROMETHEE-2
and TOPSIS for ranking the interview candidates on the basis of some experts opinion
followed by the comparison between the results and final ranking by group decision analysis
method. This three step proposed methodology is effectively more powerful than the
traditional methods.
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
140
2. MCDM Methods
2.1. Fuzzy Sets
A fuzzy set is a class of objects with a range of grades of membership. A membership
function characterizes such a set which assigns to each object a grade of membership ranging
between zero and one [5]. Fuzzy logic is a powerful mathematical tool for representing
uncertainty in every field. Their role is significant when applied to complex phenomena
which are not easily described by traditional mathematical methods, especially when the goal
is to find a good approximation solution [6]. Fuzzy sets have proven to be an eminent way for
solving the decision problems where the information available is subjective and vague [7].
2.2. Linguistic Variable
A variable which can be a word or a sentence in a natural or artificial language is often
referred to as linguistic variable [8]. For example, marks can be a linguistic variable if they
are assumed to be the fuzzy variables labeled very good, good, bad, very bad etc rather than
numbers 0,1,2,3,4 For some situations where the evaluation becomes too complex and
conventional quantitative terms cannot be used, linguistic variables can provide a means of
approximate characterization of the phenomena. The main applications of the linguistic
approach lie especially in the fields of artificial intelligence, linguistics, human decision
processes, pattern recognition, psychology, law, medical diagnosis, information retrieval,
economics and related areas [8].
2.3. Fuzzy Numbers
A fuzzy numberM
~
is a convex normalized fuzzy set M
~
of the real line R which exists [7]
such that one
o
x R with 1 ) (
=
o
M
x (
o
x is called mean value of M
~
) and ) (
o
M
x is
piecewise continuous. Triangular fuzzy numbers (TFNs) are often convenient to work with
because of their computational simplicity, and they are useful in representation and
information processing in a fuzzy environment. In this study TFNs are adopted in the fuzzy
AHP method. TFNs can be defined as a triplet (l, m, u). The parameters l, m, u indicate the
smallest possible value, the most promising value, and the largest possible value that
describes a fuzzy event. A triangular fuzzy number M
~
is shown in Figure 1:
Figure 1. Triangular Fuzzy Number
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
141
2.4. Fuzzy Analytic Hierarchy Process
Fuzzy Analytic Hierarchy Process (FAHP) is an extension of Analytic Hierarchy Process
(AHP) in fuzzy environment [9]. In case of AHP, elements of pair wise comparison matrix lie
between 1 and 9, whereas in the case of fuzzy AHP these are fuzzy numbers. In this study the
fuzzy AHP is implemented, which was originally introduced by Chang [10].
Let { }
n
x x x x X ...... , ,
3 2 1
= an object set and { }
n
g g g g G ...... , ,
3 2 1
= be a goal set.
According to Changs extent analysis each object is taken from the pair wise comparison
matrix. Then, extent analysis for each goal is performed, respectively. Therefore, m extent
analysis values for each object can be obtained with the following signs:
m
gi gi gi gi
M M M M ,..... , ,
3 2 1
n i ,.... 3 , 2 , 1 =
where
j
gi
M ( j =1,2,3m ) all are TFNs. The steps of Chang [10] extent analysis can be
given as in the following:
Step 1: The value of fuzzy synthetic extent analysis is calculated as :
1
1 1 1
= = =
(
=
n
i
m
j
j
gi
m
j
j
gi i
M M S
(1)
Step 2: If
2 1
,M M be two triangular fuzzy numbers, the degree of possibility of
1 2
M M is ( )
1 2
M M V defined as
| | )) (
~
), (
~
min( sup ) (
2 1 1 2
y M x M M M V = (2)
And can be expressed as follows:
( )
( )
( ) ( )
= = =
otherwise
l m u m
u l
u l if
m m if
d M M M hgt M M V
1 1 2 2
2 1
2 1
1 2
2 2 1 1 2
, 0
, 1
) (
~ ~ ~
) (
(3)
Figure 2 illustrates Eq (3) where d is the ordinate of the highest intersection point D
between
1
M and
2
M . To compare ) , , (
1 1 1 1
u m l M = and ) , , (
2 2 2 2
u m l M = , we need both the
values of ) (
1 2
M M V and ) (
2 1
M M V .
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
142
Figure 2. Degree of Possibility
Step 3: The degree possibility for a convex fuzzy number to be greater than k convex
fuzzy ) ,.... 2 , 1 ( k i M
i
=
numbers can be defined by
( ) k i M M V M M M M V
i k
,.... 3 , 2 , 1 , min( ) ,..., , (
2 1
= = (5)
Assuming,
) ( min ) (
k i i
S S V A d =
for k =1, 2,n; k i
Then the weight vector is given by
( )
T
n
A d A d A d W ) ( ),...., ( ), (
2 1
= (6)
where ) .... 3 , 2 , 1 ( n i A
i
= are the n elements.
Step 4: Repeat steps 1 to 3 for all criteria under each experts individually.
Step 5: Calculate the geometric mean of each cell of each expert for each alternative and
each criteria to obtain the final decision matrix.
Step 6: Obtained decision matrix is normalized using Eq. (7):
ij
J
j
ij
ij
w
w
r
=
=1
j=1,2,.,J i=1,2,...n (7)
where J is the no. of alternatives and n is the no. of criteria.
ij
w is the value in each cell of
the decision matrix.
Step 7: Weighted normalized decision matrix is formed using Eq. 8:
ij ij ij
w r v = ; j=1,2,.,J i=1,2,...n (8)
2.5. PROMETHEE-2
PROMETHEE-2 (Preference Ranking Organisation METHod of Enrichment Evaluation)
is a MCDM method of outranking nature. It is one among the 5 versions of PROMETHEE,
i.e., 1 to 5 [12, 13]. The method is based on preference function approach [14]. Thus it
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
143
considers a preference between alternatives individually. In this study, PROMETHEE-2 is
used to obtain the final ranking from the weighted normalized matrix.
The steps of this method are as follows:
The input of PROMETHEE is the weighted normalized matrix which is obtained from
Fuzzy-AHP.
Step 1: Normalize all the values according to any normalization method.
Determine the deviation based on pair wise comparisons of each alternative k and l , within
each criteria i is calculated,
li ki i
r r l k d = ) , ( (9) l k, =1,2...J and i =1,2.n
where, ) , ( l k d
i
denotes the difference between the evaluations of alternatives k and l on
each criterion,
ki
r denotes the normalized value of k
th
criteria and i
th
alternative.
Step 2: Calculate the threshold values of each criteria by Eq. (10) :
2 , 1 , ,..., 2 , 1
) 1 (
) , (
,
1 ,
= =
=
=
=
l k n i
J J
l k d
J l k
l k
l k
i
i
(10)
Step 3: Calculate the preference function using any preference function, in this paper
Gaussian function is used as a preference function, ) (d P
i
for each criteria i:
>
|
|
.
|
\
|
=
0 ) , (
2
exp 1
0 ) , ( 0
) , (
2
2
l k d if
d
l k d if
l k P
i
i
i
i
(11)
Step 4: Calculate the preference index and constitute the preference index matrix using Eq.
(12):
) , ( ) , (
1
l k P w l k
i
n
i
i
=
=
l k, J (12)
Thus, we will be getting an Alternative Vs Alternative matrix whose diagonal elements are
0 as the values show the preference of alternative
k
A over alternative
i
A .
i
w
is the weight of
each criteria.
Step 5: Calculate the net outgoing flow, this represents the strength of an alternative over
other alternatives. The outgoing flow is calculated as:
=
+
=
J
j k
k
j
k j
J
1
) , (
1
1
j, k=1,2,,J (13)
Step 6: Calculate the net entering flow, which represents the weakness of an alternative
over other alternatives.
=
J
j k
k
j
j k
J
1
) , (
1
1
j, k=1,2,,J (14)
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
144
Step 7: Calculate net flow
net
j
based on the difference between outgoing and entering
flows of alternative j using Eq. (15):
+
=
j j
net
j
j=1,2,J (15)
Ranking of the alternatives is done according to the
net
j
values. Thus, the alternative that
has the highest net flow is preferable.
The following steps of Promethee-2 can be implemented in C-Code as follows:
/* PROMETHEE-2 */
#include<stdio.h>
#include<math.h>
#include<stdlib.h>
int main(){
float DecMat[10][10], Dev[30][10], Thres[10], Pref[30][10], PI[10][10], Netflow[10],
wei[10], RowSum[10];
int i,j,alt,cri;
printf("Enter no. of criteria");
scanf("%d",&cri);
printf("\nEnter no. of alternative");
scanf("%d",&alt);
printf("\nEnter the weighted normalized matrix(row wise)\n");
for(i=0;i<alt;i++){
for(j=0;j<cri;j++)
scanf("%f",&DecMat[j][i]);
}
printf("\nThe given matrix is:\n");
for(i=0;i<alt;i++){
for(j=0;j<cri;j++)
printf(" %f",DecMat[i][j]); //The weighted normalized matrix in transposed form
printf("\n");
}
float sum=0.0; //Normalizing the weighted normalized matrix
for(i=0;i<cri;i++){
for(j=0;j<alt;j++){
sum=sum+DecMat[i][j];
}
RowSum[i]=sum; //Calculating Sum of each row before normalization
sum=0.0;
}
printf("\nThe row summation is\n");
for(i=0;i<cri;i++)
printf(" %f",RowSum[i]);
for(i=0;i<cri;i++){
for(j=0;j<alt;j++)
DecMat[i][j]=(DecMat[i][j]/RowSum[i]); //Normalization method
}
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
145
printf("\nThe normalized matrix of the weighted normalized matirx\n\n");
for(i=0;i<cri;i++){
for(j=0;j<alt;j++)
printf(" %f",DecMat[i][j]);
printf("\n"); }
printf("\nEnter the weights of the criteria\n"); //Weights of the criteria calculated by FAHP
for(i=0;i<cri;i++)
scanf("%f",&wei[i]);
int m=0,k=0;
for(k=0;k<cri;k++){ //The Deviation amplitude matrix
for(i=0;i<alt;i++){
for(j=0;j<alt;j++){
if(i==j)
continue;
else{
Dev[m][k]=(DecMat[k][i]-DecMat[k][j]);
//Calculating Deviation
m=m+1;
}
}
}
m=0;
}
m=alt*(alt-1);
printf("\nThe Deviation Matrix is \n");
for(j=0;j<m;j++){
for(i=0;i<cri;i++)
printf(" %f",Dev[j][i]);
printf("\n");
}
sum=0.0;
for(i=0;i<cri;i++){ //Calculating Threshold Values of each criteria
for(j=0;j<m;j++){
sum=sum+fabs(Dev[j][i]);
Thres[i]=sum/((float)(cri*(cri-1)));
}
sum=0.0;
}
printf("\nThe Threshold values are \n");
for(i=0;i<cri;i++)
printf("\n%f",Thres[i]);
for(i=0;i<cri;i++){
//Preference Function matrix
for(j=0;j<m;j++){
if(Dev[j][i]<=0)
Pref[j][i]=0;
else
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
146
Pref[j][i]=(1-exp((-(Dev[j][i]*Dev[j][i]))/(2*(Thres[i]*Thres[i])))); //Gaussian Preference
Function
}
}
printf("\nThe Preference Function is \n");
for(i=0;i<m;i++){
for(j=0;j<cri;j++)
printf(" %f",Pref[i][j]);
printf("\n");
}
int p;
p=k=0;
sum=0.0;
for(i=0;i<m;i++){
//Preference Index Matrix
for(j=0;j<cri;j++){
if(p==k){
PI[p][k]=0.0; //Diagonal elements of Prefence Index Matrix is 0
k=k+1;
}
sum=sum+(Pref[i][j]*wei[j]);
}
PI[p][k]=sum;
k=k+1;
sum=0.0;
if(k==alt){
p=p+1;
k=0;
}
PI[alt-1][alt-1]=0.0;
printf("\nPreference Matrix is\n");
for(i=0;i<alt;i++){ //Preference Index is Alternative Vs Alternative Matrix
for(j=0;j<alt;j++)
printf(" %f",PI[i][j]);
printf("\n");
}
float inflow=0.0,outflow=0.0;
for(i=0;i<alt;i++){
for(j=0;j<alt;j++){
outflow=outflow+PI[i][j];
inflow=inflow+PI[j][i];
}
outflow=outflow/4; // Outgoing flow
inflow=inflow/4; // Incoming flow
Netflow[i]=outflow-inflow; //NetFlow =Outgoing flow - Incoming flow
outflow=inflow=0.0;
}
printf("\nThe Netflow is \n");
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
147
for(i=0;i<alt;i++)
printf("\nS%d - %f\n",i,Netflow[i]);
return(0);
}
2.6. TOPSIS
TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) is one of the
useful Multi Attribute Decision Making techniques that are very simple and easy to
implement. Unlike AHP the user need not to have a detailed knowledge about the criteria in
the decision hierarchy to make informed decisions [15]. TOPSIS method was firstly proposed
by Hwang and Yoon [16]. According to this technique, the best alternative would be the one
that is nearest to the positive ideal solution and farthest from the negative ideal solution. The
positive ideal solution maximizes the benefit criteria and minimizes the cost criteria, whereas
a negative ideal solution maximizes the cost criteria and minimizes the benefit criteria [17]. In
this study, TOPSIS method is used to determine the final ranking. The steps of this method
are as follows:
The input of TOPSIS is the weighted normalized matrix which is obtained from Fuzzy-
AHP.
Step 1: Determine the Positive Ideal Solution (PIS) and Negative Ideal Solution (NIS) as
for each criteria:
{ }
=
n
v v v A ,....., ,
2 1
(16)
Where
*
n
v
gives the maximum value of n
th
criteria.
{ }
=
n
v v v A ,....... ,
2 1
(17)
Where
n
v gives the minimum value of n
th
criteria.
Step 2: Calculate the distance of each alternative from PIS and NIS as:
( )
=
=
n
j
j ij i
v v d
1
2
* *
i =1,2,3,.,J (18)
( )
=
=
n
j
j ij i
v v d
1
2
i =1,2,3,.,J (19)
where there are J alternatives and n criteria.
Step 3: The closeness coefficient of each alternative is:
+
=
i i
i
i
d d
d
CC
i=1,2,3,., J (20)
Step 4: Rank the alternatives according to the decreasing order of the
i
CC values. The
best alternative is the one with the greatest closeness to ideal solution.
The above steps of Topsis can be implemented in C- Code as follows :
/*TOPSIS*/
#include<stdio.h>
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
148
#include<math.h>
float MAXI(float DecMat[10][10],int alt, int pos){ //Calculates Positive Ideal Solution
int i; float max;
max=DecMat[pos][0];
for(i=1;i<alt;i++){
if(DecMat[pos][i]>max)
max=DecMat[pos][i];
}
return(max);
}
float MINI(float DecMat[10][10],int alt, int pos){ //Calculates Negative Ideal Solution
int i; float min;
min=DecMat[pos][0];
for(i=1;i<alt;i++){
if(DecMat[pos][i]<min)
min=DecMat[pos][i];
}
return(min);
}
int main(){
float DecMat[10][10],MaxArr[10],MinArr[10],Maxdist[10],Mindist[10],CC[10];
int i,j,alt,cri;
printf("Enter no. of criteria");
scanf("%d",&cri);
printf("\nEnter no. of alternative");
scanf("%d",&alt);
printf("\nEnter the weighted normalized matrix(row wise)");
for(i=0;i<alt;i++){
for(j=0;j<cri;j++)
scanf("%f",&DecMat[j][i]);
}
for(i=0;i<cri;i++){
MaxArr[i]=MAXI(DecMat,alt,i); //Positive Ideal Solution for each criteria
}
for(i=0;i<cri;i++){
MinArr[i]=MINI(DecMat,alt,i); //Negative Ideal Solution for each criteria
}
printf("\nThe positive and negative ideal solutions are\n");
printf("\tPIS\t\t\tNIS\n");
for(i=0;i<cri;i++)
printf(" %f\t\t %f\n",MaxArr[i],MinArr[i]);
float sum=0;
for(i=0;i<alt;i++){
for(j=0;j<cri;j++){
sum=sum+pow((DecMat[j][i]-MaxArr[j]),2); //Distance from positive ideal solution
Maxdist[i]=sqrt(sum);
}
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
149
sum=0;
}
for(i=0;i<alt;i++){
for(j=0;j<cri;j++){
sum=sum+pow((DecMat[j][i]-MinArr[j]),2); //Distance from negative ideal solution
Mindist[i]=sqrt(sum);
}
sum=0;
}
for(i=0;i<alt;i++){
CC[i]=(Mindist[i]/(Maxdist[i]+Mindist[i])); //Relative Closeness of each alternative
}
printf("\nThe relative closeness of each alternative is\n");
for(i=0;i<alt;i++){
printf("S%d - %f",i,CC[i]);
printf("\n");
}
return(0);
}
2.7. Correlation Coefficients
Assessing the correlation/consistency between different ranking patterns obtained by
different MCDM methods and/or different decision makers and/or different scenarios for a
given set of alternatives forms a major part in their comparative study. Correlation
coefficients measure the extent to which the ranks are correlated which can be used in this
regard. These values vary from +1.00 (perfect positive relation) to -1.00 (perfect negative
relationship) and value of zero indicates no relationship.
A non parametric rank correlation method namely, Spearman which is used to compute
correlation coefficient values is explained as follows:
2.7.1. Spearman Rank Correlation Method
Spearman rank correlation coefficient
is useful to determine the measure of
association/correlation (including positive or negative direction of a relationship) between
ranks achieved by different MCDM methods and/or different decision-makers and/or
different scenarios for a given set of alternatives. If Ua and Va denote the ranks achieved by
above situation(s) for the same alternative a, then R is defined as [18] :
n n
d
R
n
a
a
=
=
3
1
2
6
1
,where d
a
=difference between ranks U
a
and V
a
achieved by the same alternative a
n =number of alternatives and . 1 1 R
Various critical values for Spearman rank correlation coefficient for various significance
levels is provided in [19]. Numerous case studies have used the Spearman rank correlation
method for computation of correlation coefficient values [20].
Characteristics of R can be explained in Table 1 as:
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
150
Table 1. Characteristics of Co-efficient R
Correlation
Nature of correlation Remark
0.9 - 1.0 Very High Very Strong relationship
0.7 0.9 High Marked relationship
0.4 0.7 Moderate Substantial relationship
0.2 0.4 Low Definite relationship
< 0.2 Slight Small relationship
2.8. Group Decision Making
The real life decisions are particularly complex in nature, with personal interests and
conflicting preferences among a good number of decision-makers involved, which may lead
to an unsatisfactory conclusion and sometimes may be even erroneous. In this regard,
effective group decision-making can be viewed as a process in which individual interests are
reduced and integrated so as to form a single group preference or consensus [21]. An
integrated approach can be used where rank coefficient values can be used as a benchmark to
ascertain the conflicting nature of decision makers and their strength. The two main group
decision making methods are as follows [22]:
2.8.1. Additive Ranking Rule
The Additive Ranking rule is applied to integrate the rankings of a group of decision
makers or different MCDM methods into a consensus. A group evaluation of an alternative a
in additive ranking rule is the arithmetic mean of the rankings made by G MCDM
methods/decision makers, i.e.,
G
r w
r
G
DM
DM a DM
G
a
=
=
1
.
where G =Number of methods;
wDM
=Relative influence of each decision method DM
on expected outcome;
r DM a.
=Rank obtained for each alternative a by decision method DM;
raG =Rank obtained for each alternative a by the group of G decision methods.
2.8.2. Multiplicative Ranking Rule
A group evaluation of an alternative a in multiplicative ranking rule is the product of the
rankings made by G decision makers or different MCDM methods raised to the power 1/G.
G
G
DM
DM a DM
G
a
r w r
/ 1
1
. (
=
=
where G =Number of methods;
wDM
=Relative influence of each decision method DM
on expected outcome;
r DM a.
=Rank obtained for each alternative a by decision method DM;
raG =Rank obtained for each alternative a by the group of G decision methods.
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
151
3. Case Study
In this section, the above explained methodologies are applied to a case study, in
order to prove its applicability and validity. The study is based on a faculty interview in
an engineering college. A triplet of decision makers (Expert 1, Expert 2, Expert 3) were
asked to evaluate the set of five candidates (S1, S2, S3, S4, S5). The experts mark the
alternatives using linguistic variables (E, VG, G, A, B) which are later converted to
triangular fuzzy numbers. This evaluation was done on the basis of five attributes,
Academic Qualification (AQ), Knowledge (K), Teaching Ability (TA), Research (R)
and Presentation (P). According to the requirement of the college the candidates were
evaluated. The panelists marked the candidates using linguistic variables. The table
containing the remarks for each candidate is shown in the Table 2.
Table 2. Candidates Mark using Linguistic Variables
Academic
Qualification
(AQ)
Knowledge
(K)
Teaching
Ability (TA)
Research (R)
Presentation
(P)
Candidate Ex-1
Ex
-2
Ex
-3
Ex
-1
Ex
-2
Ex
-3
Ex
-1
Ex
-2
Ex
-3
Ex
-1
Ex
-2
Ex
-3
Ex
-1
Ex
-2
Ex
-3
S1 E V
E V
G G G G G E V
V
V
G G
S2 VG G G E V
V
V
G V
G G A E V
V
S3 E E E V
G E G A A V
V
E V
V
G
S4 E V
V
G A G A B G V
G A A A B
S5 G A G G G V
V
G V
A B G G A A
3.1. Application with Fuzzy AHP Method:
Experts use the linguistic variables, to evaluate the ratings of alternatives with respect to
each criterion and then each variable is converted into triangular fuzzy numbers. The fuzzy
numbers are defined in Table 3:
Tabl e 3. Li ngui sti c Vari abl e and Tri angul ar Fuzzy Numbers
Criteria AQ K TA R P ---
Linguistic
Variables
Strong Very Strong Absolute Weak Average Equal
Fuzzy Numbers 4,6,7 5,7,9 7,8,9 2,3,4 3,5,7 1,1,1
Grades G VG E B A ---
According to the markings of the experts, the linguistic variables are converted into
triangular fuzzy numbers and a pair wise comparison matrix is given in Table 4:
Table 4. Pair Wise Comparison Matrix
Criteria
Academic
Qualification
Knowledge Teaching Ability Research Presentation
Academic
Qualification
1,1,1 0.457,0.857,1.25 0.567,0.75,0.967 1.2,2,2.4 1,1.2,1.5
Knowledge 0.795,1.167,2.18 1,1,1 0.6,0.875,1.078 1.433,2.333,2.93 1,1.4,2
Teaching
Ability
1.034,1.333,1.76 0.928,1.149,1.66 1,1,1 2.4,2.667,3.1 1.5,1.6,1.9
Research 0.416,0.5,0.833 0.341,0.428,0.69 0.323,0.374,0.41 1,1,1 0.3,0.6,1.2
Presentation 0.667,0.833,1 0.5,0.714,1 0.526,0.625,0.66 0.833,1.667,3.33 1,1,1
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
152
Following all above mentioned steps from 1 to 7 of Fuzzy Analytic Hierarchy Process, the
individual weights of each criteria and weighted normalized matrix is calculated and given in
Table 5:
Table 5. Weighted Normalized Matrix
Weight 0.197460966 0.255012403 0.291842988 0.088486794 0.167196848
Candidate
Academic
Qualification
Knowledge
Teaching
Ability
Research Presentation
S1 0.207868425 0.141985995 0.190900948 0.277981878 0.171561269
S2 0.13391749 0.229873928 0.264506429 0.097644208 0.29949027
S3 0.232906151 0.178280943 0.113804966 0.274378751 0.188175683
S4 0.204980087 0.098380755 0.02002437 0.107789208 0.023749001
S5 0.099143279 0.155196772 0.264506429 0.00736647 0.077250077
3.2. Application with PROMETHEE -2 and TOPSIS Method:
In this section, the above mentioned steps of PROMETHEE 2 and TOPSIS are applied to
the weighted normalized matrix found in Table 4 and individual ranks of each alternative by
both the methods is given in Table 6:
Table 6. Ranking of Candidates according PROMETHEE-2 and TOPSIS
Candidate Rank
PROMETHEE-2 TOPSIS
S1 3 1
S2 1 2
S3 2 3
S4 5 5
S5 4 4
3.3 Correlation Coefficients
The ranks determined by Promethee-2 and Topsis differ from each other as given in Table
7. Thus, to determine the measure of association between the ranks achieved by these two
methods, the Spearman Rank Correlation Method as mentioned in Section 2.7.1 is used to
find the correlation coefficient
.
Tabl e 7. Di fference between PROMETHEE-2 and TOPSIS
Candidate
Rank Difference in
ranks
PROMETHEE-2 TOPSIS
S1 3 1 2 4
S2 1 2 -1 1
S3 2 3 -1 1
S4 5 5 0 0
S5 4 4 0 0
da
2
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
153
From Table-7,
7 . 0 = R
establishes a Marked Relationship between Promethee-2
&Topsis.
3.4. Group Decision Making
Table 8 presents the group decision making analysis following the rules given in Section
2.8.
Table 8. Additive and Multiplicative Ranking
Alternatives
Rank
Additive
Ranking
Inferred
Group
Ranking
Multiplicative
Ranking
Inferred
Group
Ranking
Promethee-2 Topsis
S1 3 1 2.00 2 1.732 2
S2 1 2 1.50 1 1.414 1
S3 2 3 2.50 3 2.449 3
S4 5 5 5.00 5 5.000 5
S5 4 4 4.00 4 4.000 4
From Table-8, the final ranking can be given as:
Table 9. Final Ranking of the Candidates
Alternatives Rank
S1 2
S2 1
S3 3
S4 5
S5 4
4. Conclusion
In this case study of the faculty selection interview in an Engineering Organization the
uncertainty and vagueness of the experts marks have been effectively represented and
resolved to a more effective decision using the Fuzzy AHP MCDM method. For a
comparative study the result of Fuzzy AHP was separately used as an input to PROMETHEE-
2 and TOPSIS. Since, the final ranking of these two methods slightly differs from each other;
a group decision making method is implemented to get a single ranking structure. There is no
doubt for the 4th and 5th positions taken by S5 and S4 respectively but S1, S2 and S3 all are
strong contenders for the best position. S1 and S3, despite having a very strong Academic
Qualification and Research work are not chosen as the best choice as they are significantly
weak in skill of Teaching Ability (TA) than S2. In comparison to this, S2 which although has
comparatively not so strong Academic Qualification and Research work but is very strong in
depth of Knowledge and Teaching Ability, which are more important criteria in an
Engineering institute than others. Hence, S2 is the best choice for a faculty member in this
Engineering Organization.
Analyzing the importance of each criterion in Engineering Institutes and comparing the
ranking pattern of individual MCDM methods with the final ranking it can be concluded that
TOPSIS and PROMETHEE- 2 share a marked relationship between them as established by
Spearman Rank Correlation Method and the difference between them exists as TOPSIS
compares the alternatives on the overall marking of the experts for each criterion and
PROMETHEE-2 compares the alternatives on the basis of the priority of the criteria.
International Journal of u- and e- Service, Science and Technology
Vol. 6, No. 4, August, 2013
154
References
[1] P. Kumar Dey, S. Chattaraj and D. Nath Ghosh, Faculty Selection in Engineering Organization using AHP
& TOPSIS, International J ournal of Information and Computing Science (IJ ICS) 06/2012, vol. 15, no. 1,
(2012).
[2] P.-C. Chen, A Fuzzy Multiple Criteria Decision Making Model in Employment Recruitment, IJ CSNS
International J ournal of Computer Science and Network Security, vol. 9, no. 7, (2009) J uly, pp. 113-117.
[3] K. Oraee and E. Bakhtavar, Selection of Tunnel Support Systemby Using Multi Criteria Decision-Making
Tools, 29th International Conference on Ground Control in Mining.
[4] D. Chatterjee and B. Mukherjee, Study of Fuzzy-AHP Model To Search The Criterion in the Evaluation of
the Best Technical Institutions: A Case Study, International J ournal of Engineering Science and Technology,
vol. 2, no. 7, (2010), pp. 2499-2510.
[5] L. Z. Zadeh, Fuzzy sets, Inf. Control., vol. 8, (1965), pp. 338-353.
[6] G. Bojadziev and M. Bojadziev, Fuzzy sets and fuzzy logic applications, World Scientific, Singapore,
(1998).
[7] H. J . Zimmermann, Fuzzy set theory and its applications, Kluwer, Boston, (1992).
[8] L. A. Zadeh, The concept of a linguistic variable and its application to approximate reasoning-I, Inf.
Science, (1975), pp. 199-249.
[9] T. L. Saaty, How to Make a Decision: The Analytic Hierarchy Process, European J ournal of Operational
Research, vol. 48, (1990), pp. 9-26.
[10] D. Y. Chang, Applications of the Extent Analysis Method on Fuzzy AHP, European J ournal of Operational
Research, vol. 95, (1996), pp. 649-655.
[11] C. Kahraman, U. Cebeci and D. Ruan, Multi-Attribute Comparison of Catering Service Companies Using
Fuzzy AHP: The case of Turkey, International Journal of Production Economics, vol. 87, no. 2, (2004), pp.
171-184.
[12] M. F. A. Taleb and B. Mareschal, Water Resources Planning in the Middle East: Application of the
PROMETHEE V Multicriteria Method, European J ournal of Operational Research, vol. 81, (1995), pp. 500-
511.
[13] J . C. Pomerol and S. B. Romero, Multicriterion Decision in Management: Principles and Practice, Kluwer
Academic, Netherlands, (2000).
[14] J . P. Brans, P. Vincke and B. Mareschal, How to Select and How to Rank Projects: The PROMETHEE
method, European J ournal of Operational Research, vol. 24, (1986), pp. 228-238.
[15] D. Nath Ghosh, Analytic Hierarchy Process and TOPSIS Method to Evaluate Faculty Performance in
Engineering Education, UNIASCIT, vol. 1, no. 2, (2011), pp. 63-70.
[16] C. L. Hwang and K. Yoon, Multiple Attribute Decision Making Methods and Applications, Springer,
Berlin Heidelberg, (1981).
[17] Y. J . Wang and H. S. Lee, Generalizing TOPSIS for fuzzy multiple-criteria group decision making,
Computers and Mathematics with Applications, vol. 53, no. 11, (2007), pp. 1762-1772.
[18] C. Spearman, The Proof and Management of Association between Two Things, American J ournal of
Psychology, vol. 15, (1904), pp. 72-101.
[19] G. Woodbury, Introduction to Statistics, Thomson Learning, USA, (2002).
[20] K. S. Raju and D. Nagesh Kumar, Multi-criterion Decision-Making in Irrigation Planning, Agricultural
Systems, vol. 62, no. 2, (1999), pp. 117- 129.
[21] P. H. Liu and C. C. Wei, A Group Decision-Making Method for Appraising the Performance of
Organizations, International J ournal of the Computer, the Internet and Management, vol. 8, no. 2, (2000), pp.
39-49.
[22] T. X. Bui, Coop: A Group Decision Support System for Cooperative Multiple Criteria Group Decision-
making, Springer-Verlag, Berlin, (1987).
[23] K. S. Raju and D. Nagesh Kumar, Multicriterion Analysis in Engineering and Management, PHI Learning
Pvt. Ltd., (2010).
