Bab 4: Hasil dan Pembahasan
38 Sebanyak 40
CSFs
telah teridentifikasi melalui kajian pustaka mereferensi pada
Lampiran 1
. Selanjutnya didisain kuisioner dengan lima skala Likert menggunakan 40 variabel
CSFs
tersebut. Sebelum disebarkan kepada responden, kuisioner tersebut terlebih dahulu telah didiskusikan dengen 5 orang para pakar ahli dan profesional dibidang
managemen biaya proyek konstruksi.
4.3. Analisis Prioritas CSFs dengan Metoda AHP
The most important CSFs are analysed using AHP techniques. The three main stages of AHP techniques are implemented during data analysis refer to Jaya and Pathirage, 2013.
Firstly, calculating the relative importance between four evaluation-criteria i.e., Project Type, Project Phase, Project Monitoring, and Project Deliveries in respect of the
decision-goal the most important CSFs; Second, calculating the relative importance between eight CSF-alternatives i.e.,
MARCON, DEVFOC, INVTEC, LOCRES, INTBEN, PROCOM, ROPRAC, and METOOL in respect of each element of the four different evaluation-criteria;
Third, determining the relative importance ranking of eight CSF-alternatives under the four evaluation-criteria through establishing the AHP solution tree for the highest
priority CSFs.
4.4. The Relative Importance of the Evaluation-Criteria
The judgements of five experts were considered to determine the relative importance of four evaluation-criteria using five Likert scales whilst scoring forty variable
CSFs. The cumulative scores of every element of the four evaluation-criteria were used to develop pair-wise comparisons between them. The relative importance of each element of
evaluation-criteria over another was analysed in respect of the decision-goal as structured on the top level of Figure 3-3. AHP can be expressed by examining pair-wise comparisons
to develop matrix algebra, and squaring the matrix with multiple iterations until the last normalised eigenvector does not change too much to the preceding iteration. Prescribed
measures are placed in four digit decimals of change values between consecutive iterated
Bab 4: Hasil dan Pembahasan
39 �
…
… ⋮ ⋮ ⋮ ⋮
… matrices e.g., Haas and Meixner, 2005; and Hamdeh, 2010. However, greater decimals
would provide more precise calculations for better results.
4.4.1. Pair-Wise Comparisons of the Evaluation-Criteria
The dimension of pair-wise comparisons is calculated by given
Equation 1
. The eigenvalue methods
Equation 2
and its reciprocal values are used to form the matrix
A
through comparing each element of evaluation-criteria Project Type, Project Phase, Project Monitoring, and Project Deliveries. Following the matrix formula
Equation 3
, it is deconstructed as the matrix algebra in Table 4-1.
Number of Comparisons � � =
× −
Equation 1 Eigenvalue method:
� − � �� × � = Equation 2
Matrix formula:
Equation 3
Where, Dimension of matrix A
� Pair-wise comparison matrix A
� � Principal eigenvalue of matrix A
� Relative weights eigenvectors of matrix A
, , ..., and
The weights of element 1, 2, ..., and n The
� � and eigenvalue are calculated using Equation 1 and 2 to decompose the matrix algebra A as explained in the example below:
Bab 4: Hasil dan Pembahasan
40 � � =
× −1
2
=
4 ×4 −1
2
= 6
= 725
778 = 0.931877
Where, � � = 6
The 6 sets of pair-wise comparisons are located on the top-right side of diagonal matrix A and its reciprocals placed on bottom-left side
= 725 The weight of element number 1 Project Type
= 778 The weight of element number 2 Project Phase
0.931877 The eigenvalue of Project Type relative to Project Phase, included in
the matrix algebra The eigenvalues e.g., 0.91877, 0.834292, etc. and their reciprocals 1.073103,
1.198621, etc. of each element relative to other elements are represented in Table 4-1.
Tabel 4-1: Matrix Algebra of the Evaluation-Criteria
4.4.2. Eigenvector Solution of the Evaluation-Criteria
The process of AHP techniques must be iterated properly in order to achieve the best result on the eigenvector solution. The following matrices in Table 4-2 and Table 4-3
are the last two squared and iterated matrices, and normalised eigenvectors are calculated by dividing each row sum with the total column sum.
Criteria Types
Phases Monitors
Deliveries Types
1 0.931877
0.834292 1.186579
Phases 1.073103
1 0.895282
1.273322
Monitors 1.198621
1.116967 1
1.422259
Deliveries 0.842759
0.785347 0.703107
1 Column Sum
4.114483 3.834190
3.432681 4.882160
Bab 4: Hasil dan Pembahasan
41
Tabel 4-2: The First Iteration of Squared Matrices of the Evaluation-Criteria
The normalised eigenvector summation, for example: all elements of Evaluation- criteria Project Type, Project Phase, Project Monitoring, and Project Delivery should
always be an absolute value of
‘one’
,
as shown in Table 4-2 and Table 4-3.
Tabel 4-3: The Second-Last Iteration of Squared Matrices of the Evaluation-criteria
There is not a different value between each normalised eigenvector in the first iteration matrix Table 4-2 and the second-last iteration Table 4-3. It means that the
latest iteration matrix Table 4-3 has provided the best result for the eigenvector solution.
4.4.3. Consistency Ratio of the Evaluation-Criteria
The level of inconsistency must be checked to see if the result relating to the relative importance of evaluation-criteria was derived from acceptable consistent matrices.
If the outcomes are considered to be valid, robust enough and makes sense, it should be obtained from consistent or near consistent matrices Ishizaka and Labib, 2009. The
principal eigenvalue is necessary for examining the level of inconsistency of the matrix Saaty, 2008.
Saaty 1977 has calculated the Consistency Index
CI
and Consistency Ratio
CR
using the given
Equation 4
and
Equation 5
below: �� =
� �� − −
Equation 4
Criteria Types
Phases Monitors
Deliveries Row Sum
Eigenvector
Types 4
3.727506 3.337169
4.746318 15.810993
0.243044 Phases
4.292414 4
3.581128 5.093290
16.966831 0.260811
Monitors 4.794483
4.467866 4
5.689034 18.951383
0.291317 Deliveries
3.371034 3.141388
2.812428 4
13.324851 0.204827
Column Sum 16.457931
15.336761 13.730725
19.528642 65.054059
1
Criteria Types
Phases Monitors
Deliveries Row Sum
Eigenvector Changes Ranking
Types 64
59.640103 53.394707
75.941080 252.975890
0.243044 0.00 3rd
Phases 68.678621
64 57.298044
81.492635 271.469299
0.260811 0.00 2nd
Monitors 76.711724
71.485861 64
91.024550 303.222135
0.291317 0.00 1st
Deliveries 53.936552
50.262211 44.998849
64 213.197612
0.204827 0.00 4th
Column Sum 263.326897
245.388175 219.691600
312.458265 1,040.864936
1 0.00
Bab 4: Hasil dan Pembahasan
42 �� =
�� ��
Equation 5 Where,
�� 10 Consistency Ratio less than 10 an acceptable inconsistency of the matrices
�� Random Index
The average random index was randomly generated from reciprocals through analysing a sample size of 500 matrices Ishizaka and Labib, 2009, and this is
provided in Table 4-4.
Tabel 4-4: Random Inconsistency Index RI
The consistency ratio has been calculated following this procedure with given
n= 4
units and
RI= 0.90
, resulting in
CR~0.00 10.
This indicates that the matrices examined are
near perfectly consistent
. Therefore, outcomes are considered valid.
4.5. The Relative Importance of the CSF-Alternatives
The individual responses of 107 project professionals were provided using five Likert scales through scoring forty variable CSFs. Cumulative weights of every factor of
eight CSFs are used to develop pair-wise comparisons. Following similar AHP procedures to that of
Section 4.4
, the relative importance of the CSF-alternatives is analysed in respect of the four elements of the evaluation-criteria as shown in Figure 1 Project Type, Project
Phase, Project Monitoring, and Project Deliveries. For a calculation example, the relative importance of CSF-alternatives under the
Project Type can be expressed by examining pair-wise comparisons through matrix algebra as provided in Table 4-5.
Bab 4: Hasil dan Pembahasan
43
Tabel 4-5: Matrix Algebra of CSF-alternatives under the Project Type
A normalised eigenvector of the iteration matrices can be obtained through squaring the matrices that are calculated in Table 4-6 and Table 4-7. These matrices are
measured in six digit decimals.
Tabel 4-6: The Before-Last Iteration of Squared Matrices of CSF-alternatives under the Project Type
Normalised eigenvectors between the last two squared and iterated matrices Table 4-6 and Table 4-7 have shown insignificant changes. The highest change is represented by
PROCOM =0.0005297. Therefore, an attempt to iterate one more squared matrix would not have changed the ranking of normalised eigenvectors. This eigenvector may be called
the best eigenvector solution refer to Table 4-7.
Tabel 4-7: The Last Iteration of Squared Matrices of CSF-alternatives under the Project Type
Alternatives MARCON DEVFOC INVTEC
LOCRES INTBEN PROCOM ROPRAC METOOL Row Sum
MARCON 1 1.213592 1.30813953 1.322751 1.470588
1.12069 1.227094 0.953895 9.61675
DEVFOC 0.824
1 1.07790698 1.089947 1.211765 0.923448 1.011126 0.78601 7.924202
INVTEC 0.764444 0.927724
1 1.01117 1.124183 0.856705 0.938045
0.7292 7.351471 LOCRES
0.756 0.917476 0.98895349 1 1.111765 0.847241 0.927683 0.721145 7.270263
INTBEN 0.68 0.825243 0.88953488 0.899471
1 0.762069 0.834424 0.648649 6.53939
PROCOM 0.892308 1.082898 1.16726297 1.180301 1.312217
1 1.094946 0.851168 8.5811
ROPRAC 0.814933 0.988997 1.06604651 1.077954 1.198431 0.913287
1 0.777361 7.83701
METOOL 1.048333 1.272249 1.37136628 1.386684 1.541667 1.174856 1.286404
1 10.08156 Column Sum 6.780019 8.228178 8.86921064 8.968279 9.970616 7.598297 8.319722 6.467427 65.20175
Alternatives MARCON DEVFOC INVTEC
LOCRES INTBEN PROCOM ROPRAC METOOL Row Sum Eigenvector
MARCON 8 9.708738 10.4651163 10.58201 11.76471 8.965517 9.816754 7.631161
76.934 0.14797563
DEVFOC 5.768
8 8.62325581 8.719577 9.694118 7.387586 8.089005 6.288076 62.56962 0.120347
INVTEC 6.115556 7.421791
8 8.089359 8.993464 6.85364 6.508201 5.833598 57.81561
0.111203
LOCRES 6.048 7.339806 7.91162791
8 8.894118 6.777931 7.421466 5.769157 58.16211 0.11186958
INTBEN 5.44 6.601942 7.11627907 7.195767
8 6.096552 6.675393 5.189189 52.31512 0.10062343
PROCOM 7.138462 8.663181 9.33810376 10.49996 10.49774
8 7.76858 6.809343 68.71537
0.13216782
ROPRAC 6.519467 7.911974 8.52837209 8.623633 9.587451 7.306299
8 6.268455 62.74565 0.12068561
METOOL 8.386667 10.17799 10.9709302 11.09347 12.33333 9.398851 10.29123
8 80.65248 0.15512778
Column Sum 53.41615 65.82543 70.9536852 72.80378 79.76493 60.78638 64.57063 51.78898 519.91
1
Alternatives MARCON DEVFOC INVTEC
LOCRES INTBEN PROCOM ROPRAC METOOL Ror Sum
Eigenvector Rank Changes
MARCON 504 621.3592 669.767442 686.7302 752.9412 573.7931 608.9626 488.8809
4906.43 0.1481122 2nd
-0.0001366
DEVFOC 408.704
504 543.265116 557.1461 610.7294 465.4179 493.6962 396.5498 3979.51
0.1201308 5th 0.0002163
INVTEC 378.7856
467.113 503.5044 516.3765
566.031 431.3547 457.5488 367.4779 3688.19
0.1113367 7th -0.0001336
LOCRES 381.024 469.7476 506.344186
519.168 569.2235 433.7876 460.3757 369.5939 3709.26
0.1119728 6th -0.0001032
INTBEN 342.72 422.5243
455.44186 466.9765 512 390.1793 414.0946
332.439 3336.38
0.1007163 8th -0.0000929
PROCOM 449.6584 554.3652 597.554099 612.6891 671.7602 511.9276 526.6293 436.1214
4360.71 0.1316381 3rd
0.0005297
ROPRAC 411.1421 506.8708 546.359623 560.1892 614.2082
468.069 496.774 398.8019
4002.41 0.1208222 4th
-0.0001366
METOOL 528.36 651.3916 702.139535 719.9221 789.3333 601.5264 638.3958 512.5101
5143.58 0.1552709 1st
-0.0001432 Column Sum
3404 4197.372 4524.37622 4639.198 5086.227 3876.056 4096.477 3302.375 33126.47 1
0.0000000
Bab 4: Hasil dan Pembahasan
44 A consistency ratio for this matrix has been checked following procedures as
Equations 4
and
Equation 5,
given
n= 8 units,
with
RI= 1.41
refer to Table 4-4. The result shows that
CR = 0.019 10
. This indicated the valid outcomes that the matrices examined are nearly consistent.
The relative importance of the CSF-alternatives and their consistency ratios under the other three evaluation-criteria i.e., Project Phase, Project Monitoring, and Project
Deliveries have been analysed through the same procedures as applied for CSF- alternatives under the Project Type. Their valid outcomes are also presented in the left-
hand side of Table 4-8.
Tabel 4-8: AHP Solution Matrix of CSF-Alternatives under Evaluation-criteria
The best eigenvector solution of evaluation-criteria is derived from Table 4-3 into the right-hand side of Table 4-8, and all of the best eigenvector solutions of CSF-
alternatives are restructured to develop an AHP Solution Tree.
4.6. The Solution through AHP Tree