3.2. Fuzzy Number
Fuzzy logic is an area that has developed significantly in recent years. The basic settings of fuzzy logic are set by Lotfi Zadeh [
45,
46,
47], and those were enough to provide further development of the area of fuzzy logic causing its wider application in the practice.
In conventional sets, the membership of one element to a set is unambiguously defined, a certain element can or cannot belong to a set. In fuzzy sets, the membership of one element to a particular set is not precisely defined; the element can more or less form part of a set, which makes fuzzy logic closer to human perception compared with conventional logic. Such a feature allows fuzzy logic to quantify information that is considered imprecise in classic logic. The existence of seemingly inaccurate information, which is well processed by fuzzy logic, is a very common occurrence in the social sciences, and thus in decision-making processes.
The quantification of uncertainty has advanced significantly in modern science. Today, in addition to classic fuzzy numbers, a variety of other fuzzy approaches are used, such as: type-2 fuzzy numbers, intuitionistic fuzzy numbers, Pythagorean fuzzy numbers, etc. In addition, other approaches are used to quantify uncertainty, such as: rough numbers, gray numbers, D numbers, etc. In this paper, the fuzzification of the applied method was carried out with classic triangular fuzzy numbers. Checking the obtained results, it was found that classic triangular fuzzy numbers described quite well the uncertainty arising when solving the decision-making problems presented in this paper. Accordingly, the following part of the paper presents the most basic elements that are sufficient to understand the application of triangular fuzzy numbers.
Fuzzy set
A is defined as a set of ordered pairs:
where:
- -
X—universal set or a set of considerations based on which is defined the fuzzy set A;
- -
μA(x) membership function of the element x (x ∈ X) to the set A; membership function can have any value between 0 and 1, and as the value of the function is closer to one, the membership of the element x to the set A is higher, and vice versa.
Triangular fuzzy number
T = (
t1,
t2,
t3) is presented in
Figure 2, where
t1 presents the left, and
t3 the right distribution of the confidence interval of the fuzzy number
T and
t2, the point in which the membership function of the fuzzy number has its maximum value.
The concept based on which the fuzzy number is expressed using confidence intervals and relevant confidence degrees is suggested by Kaufmann and Gupta [
48]. Considering
Figure 2, the confidence interval presents a closed set [
t1,
t3]. Therefore, a fuzzy variable can have the values only from the confidence interval. Defining confidence interval of every fuzzy variable is the task of the planner, and the most natural and often used solution is to adopt a confidence interval so that it matches the physical limits of the variable [
49]. If the variable has no physical origin, some of the standard ones are adopted or it is defined as an abstract confidence interval [
42]. The degree of membership is the value related to the confidence interval. In
Figure 2, the confidence interval with the membership degree α can be observed, marked as
.
3.3. Fuzzy Logarithm Methodology of Additive Weights
In order to define the weight coefficients of the criteria and select the best alternative, modified Logarithm Methodology of Additive Weights (LMAW) is used. This method in its basic form (crisp version) was presented for the first time in the paper by Pamučar et al. [
22]. In this paper, the modification (fuzzification) of the LMAW method is performed by applying triangular fuzzy numbers.
Further in the paper the steps of the fuzzy LMAW (FLMAW) are presented. The FLMAW method consists of six steps:
Step 1: Forming initial (expert) decision-making matrices (). In the first step, every expert () from the group of experts () defines one initial decision-making matrix, in which he performs the evaluation of alternatives in relation to criteria Accordingly, for every expert the matrix is obtained, where presents fuzzy value based on which the expert evaluated -th alternative by -th criterion. The evaluation is made based on quantitative indicators or based on fuzzy linguistic descriptors, depending on the type of criterion.
Step 2: Forming initial (aggregated) decision-making matrix (
). Aggregation of initial (expert) matrices into one aggregate matrix is made by applying the Bonferroni aggregator, according to the Equation (2):
where
presents averaged values obtained by applying the Bonferroni aggregator;
present stabilization parameters of the Bonferroni aggregator,
e presents the
e-th expert
,
l—the left distribution of a fuzzy number,
r—the right distribution of a fuzzy number, and
m—the value in which the membership function of a fuzzy number is equal to one. Before the aggregation, the quantification of linguistic criteria is performed.
Step 3: Normalization of the elements of the initial decision-making matrix. Applying the Equation (3), the normalized matrix
is obtained:
where
presents normalized values of the initial decision-making matrix, while
, and
, and
l presents the left distribution of a fuzzy number,
r the right distribution of a fuzzy number, and
m the value where the membership function of a fuzzy number is equal to one.
Step 4: Determining weight coefficients of criteria. In order to determine weight coefficients of the criteria, certain experts are supposed to be engaged .
Step 4.1: Prioritization of criteria. Based on the values from the predefined fuzzy linguistic scale the experts prioritize the criteria . To the criterion with higher significance is assigned the higher value from the fuzzy linguistic scale, and vice versa. This way the priority vectors are defined, for every expert separately, where presents the value from the fuzzy linguistic scale which the expert e () assigned to the criterion n.
Step 4.2: Defining the absolute fuzzy anti-ideal point (). This value is defined by a decision maker, and it presents a fuzzy number which is smaller than the smallest value from the set of all priority vectors.
Step 4.3: Defining fuzzy relation vector (
). Applying the Equation (4), the relation between the elements of the priority vector and the absolute anti-ideal point (
) is determined.
Applying the Equation (4), the relations vector of the expert e (): is obtained.
Step 4.4: Determining vectors of weight coefficients
, for every expert separately. Fuzzy values of the weight coefficients of criteria for the expert
e (
) are obtained by applying the Equation (5):
where
presents the elements of the relation vector
,
the left distribution of the fuzzy priority vector,
the right distribution of the fuzzy priority vector, and
m the value in which the membership function of the fuzzy priority vector is equal to one.
Step 4.5: Calculation of aggregated fuzzy vectors of weight coefficients
. The aggregated fuzzy vectors of the weight coefficients
are obtained by applying the Bonferroni aggregator, according to the Equation (6):
where
present stabilization parameters of the Bonferroni aggregator, while
presents the weight coefficients obtained based on the assessments of the
e-th expert
,
presents the left distribution of the fuzzy weight coefficient
,
presents the right distribution of the fuzzy weight coefficient
, a
presents the right value in which the function of the fuzzy weight coefficient
is equal to one.
Step 4.6: Calculation of the final values of weight coefficients. The calculation of the final values of weight coefficients of criteria is obtained by defuzzification, according to the Equation (7):
Step 5: Calculation of the weighted matrix (
). The elements of the weighted matrix
are obtained by applying the Equation (8):
where
while
presents the elements of normalized matrix
, while
presents the weight elements of the criteria,
l—the left distribution of a fuzzy number,
r—the right distribution of a fuzzy number, and
m the value in which the membership function of a fuzzy number is equal to one.
Step 6: Calculation of the final index for ranking alternatives (
). The final rank of the alternatives is defined based on the value of
, where the better ranked alternative is the one with the higher value of the
. The value of
is obtained by defuzzification of the value
, according to the Equation (7). The value
is calculated by applying the Equation (10):
where
presents the elements of the weighted matrix
,
l—the left distribution of a fuzzy number,
r—the right distribution of a fuzzy number, and
m the value in which the membership value of a fuzzy number is equal to one.