NAVY’s WORK UNIT
2. MATERIAL & METHODOLOGY
2.3 DEMATEL
The DEMATEL method has been applied in many fields, it extracts various uneasy factors in life (Tamura and Akazawa, 2005). The DEMATEL method can describe the interrelations between criteria and find the main criteria that present the effectiveness of a factor, so that a causal relationship can be analyzed between complex factors in a causal diagram. DEMATEL is used to test whether the model we make with several criteria, indeed has a relationship between the
criteria with each other before the criteria model is processed by the ANP method to find the weight of each criterion. The relationship between criteria or subcriteria in the DEMATEL method is illustrated on a scale of values from 0 to 4.
In the preparation of the DEMATEL method, there are several steps that must be taken (Chiu et al., 2006), include:
a. Stage 1: Create a linkage matrix directly by showing the scale of pairwise comparisons made into four levels. The results of paired comparisons will produce a direct linkage matrix.
b. Stage 2: Normalize the direct linkages matrix, where the main diagonal element is zero.
c. Stage 3: Obtain a total linkage matrix. After obtaining a normalized linkage matrix that is normalized, namely the M matrix, the total linkage matrix, where matrix I is an identity matrix.
d. Stage 4: Calculate the dispatcher group and receiver group. Some criteria with positive D-R values have a greater influence than the other criteria and are assumed to be the top priority (dispatchers). While the criteria with a negative D-R value receive an effect greater than the other criteria and are assumed to be the last priority (receiver).
The greater the D + R value of a criterion means to have more relationships with other criteria while the criteria with a smaller D + R value means having a relationship with other smaller criteria. (Seyed hosseini, Safaei and Asgharpour, 2006). While the value of D + R or Prominence shows the importance of these subcriteria (Lin and Wu, 2004).
e. Stage 5: Set the threshold value and get an impact-diagraph map. To get the right impact- diagraph map, in setting the threshold value for the level of influence can be determined by the decision maker or from an expert person by conducting a discussion (Tzeng, Chiang and Li, 2007). The impact-digraph map can be obtained by mapping the values (D + R, D-R), where the horizontal axis is the
ICMST 2019 August, 1 2019 value of D + R and the vertical axis is the value of D- R (Wu and Lee, 2007).
Figure 1. Impact-Diagraph Map 2.4 Analytic Network Process (ANP)
Analytic Network Process or ANP is a method of decision making on technical-social (socio- technical) issues based on a number of criteria (multi-criteria). ANP is a new approach to qualitative methods, which is intended to 'replace' the Analytic Hierarchy Process (AHP) method. ANP's strengths from other methodologies are its ability to help us measure and synthesize a number of factors in a hierarchy or network. There are no other methodologies that have synthesis facilities such as the ANP methodology. Meanwhile, the simplicity of its methodology makes ANP a more general methodology and easier to apply for various qualitative studies, such as decision making,
forecasting, evaluation, mapping, strategizing, resource allocation, and so on.
The important thing in building the ANP model is the existence of alternative choices and selection criteria through pairwise comparisons on the scale of importance 1-9, into the model, then the results will be in the form of choice priorities (Saaty, 2005). ANP is a development of the AHP (Analytical Hierarchy Process) methodology that is used to solve the Multi Criteria Decision Making (MCDM) problem that cannot be structured, because it involves the interaction and dependence of the top elements on the bottom element. ANP can model the system with feedback where one level may dominate and be dominated either directly or indirectly by other levels.
Kluster 1
Kluster 5
Kluster 2
Alternatif
Kluster 4
Kluster 3
Figure 2. Network Model
D-R
0,4 0,2
-0,2 -0,4
C
D
B
(10.095,0.421)
D+R
A
In creating a network on ANP there are several types of feedback that are used according to needs, each network has its own advantages and disadvantages. Some types of networks can be seen in Figure 2. ANP itself is intended to determine the relative importance of a set of activities in MCDM using Pairwise Comparison. In general, ANP is applied to the dominance of influence among stake holders or alternatives in relation to attributes or criteria. The scale of pair comparison is used in ANP as in Table 1
Table 1. Pairwise Comparison Scale
Source: according to saaty (1977; 1980) The steps that are generally carried out in this ANP are: 1) Defining the problem; 2) Define evaluation criteria; 3) Defining the weight of interest, where the scale of rating is of importance; 4) Defining the weight of dependency; 5) Define priority weights by multiplying the weight of importance and the weight of dependency.
2.4.1 Feedback Network.
Many decision problems cannot be arranged hierarchically because they involve interactions and dependencies of elements at a higher level with elements at a lower level. The level of alternative interest is not only determined based on the level of importance of the cluster but also determined based on the level of alternative interest itself. Feedback is also possible to factor in the future at this time to determine what we must do to get the desired goal
Intermediate Component (Transient State)
Source
Componen Source
Componen (Feedback Loop)
Sink Component ( Absorbing
State) Intermediate
Component (Recurrent state)
Outer Dependence
C3
C2 C1
C4
C5 Interdependence
Loop
Figure 3. Feedback Network Structure
ICMST 2019 August, 1 2019 The influence of a collection of elements given in a
component on each element in the system is symbolized by the priority vector. It can be formed a matrix that reflects the flow of influence of a
component element both with the element itself and with other elements in the network can be seen in equation 1 according at Saaty (1997).
W = C1
C2
W11
W21
Wn1
W12
W22
Wn2
W1n
W2n
Wnn
C1
e11e12….e1m 1
C1
e11e22….e2m 2
Cn
en1en2….enmn e11
e12
e1m1
e21
e22
e2m 2
en1
en2
enmn
Cn
….(1) Source: according to saaty (1977; 1980)
Wij input data in a supermatrix is called a block. The block is a matrix with arrangement as in equation 2 (Saaty, 1997).
… (2) Source: according to saaty (1977; 1980)
Equation 2 above shows how much influence one element has with other elements. to produce the priority limit of the supermatrix, then the supermatrix must be converted into a matrix where each column has a uniformity of number, so it is necessary to compare between these components according to
the of this process are known as weighted supermatrix which is stochastic.
2.4.2 Sensitivity analysis
To anticipate a change in decision caused by a change in weight, sensitivity analysis is carried out. This change occurs if you change the weight of a criterion due to an input that makes the criteria not so important or vice versa. This is useful for knowing how far these changes can affect the results obtained. Sensitivity analysis conducted is changing the weight of criteria and sub criteria by using super decision software.
In this study, the sampling technique was done by Non-Probability Sampling, which is a sampling technique that does not provide an opportunity (opportunity) for each member of the population to be sampled (Doherty, 1994) (Showkat
& Parveen, 2017). The Non-Probability Sampling technique used is using Purposive Sampling or
sampling considerations. Definition of purposive sampling is a sampling technique used by researchers if researchers have certain considerations in sampling or determining samples
for specific purposes (Doherty, 1994) (Showkat &
Parveen, 2017). the research stages as illustrated in the flow diagram.
Figure 4. Flowchart Diagram
3. RESULTS AND DISCUSSION