European Journal of Operational Research 174 (2006) 1317–1324

www.elsevier.com/locate/ejor

Short Communication

Comparing non-structural fuzzy decision support system

and analytical hierarchy process in decision-making
for construction problems
C.M. Tam *, Thomas K.L. Tong, Gerald W.C. Chiu
Department of Building and Construction, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon Tong, Hong Kong

Received 18 March 2003; accepted 14 March 2005

Available online 13 June 2005

Abstract

Analytical Hierarchy Process (AHP) is a commonly used decision-aiding tool for resolving multi-criteria decision
problems. However, users sometimes find it difficult to ensure a consistent pairwise comparison between voluminous
decisions. The cause of which is that the Consistency Ratio (CR) is produced after the evaluation process and its global
acceptance criteria is limited. When the derived ratio reports some inconsistency, it requires a long process to locate and
rectify the problem. The major aim of this study is to look for an alternative decision-aiding tool to AHP, helping to
avoid the above problem. The alternative approach proposed in this study is the Non-Structural Fuzzy Decision Sup-
port System (NSFDSS). The application of the system is illustrated with a worked example. The results generated by
NSFDSS are compared against those generated by the conventional AHP that shows the effectiveness and some unique
advantages of the proposed tool over AHP.
Ó 2005 Elsevier B.V. All rights reserved.

Keywords: Decision support systems; Multi-criteria construction problems; Non-structural fuzzy decision support system

1. Background of the research

The construction industry is characterized by continual changes, varying technologies, undesired work-
ing conditions, the involvement of numerous trades and operations which require enormous efforts to man-
age for acceptable outcomes. Owing to these, it is well-acknowledged that construction problems are often
ill-structured and multi-criteria in nature, involving many uncertainties (variables) and variations (Li and

*
Corresponding author. Tel.: +852 27887620; fax: +852 27887612.
E-mail address: bctam@cityu.edu.hk (C.M. Tam).

0377-2217/$ - see front matter Ó 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejor.2005.03.013
1318 C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324

Table 1
Summary of criticisms on AHP
Author (Year) Asserted Shortfalls of AHP
Belton and Gear (1983) and Dyer (1990) AHP suffered from the phenomenon of rank reversal
Belton and Gear (1985) AHP lacked of a firm theoretical basis
Zahir (1991) Uncertainties in the relative weights of any pairwise comparisons in AHP would
affect the resulting priorities of the decision elements
Murphy (1993) AHP suffered from limitations resulted from its application of consistency index
Paulson and Zahir (1995) Judgmental uncertainty during pairwise comparisons in AHP could lead to rank
reversals and weaken the decision makerÕs confidence on the results
Zeshui and Cuiping (1999) AHP was time-consuming and impracticable to deal with the unacceptable
consistency ratio

Love, 1998). As a result, multi-criteria decision-making techniques should be employed in solving these
problems. The common approach is the Analytical Hierarchy Process (AHP) (Paulson and Zahir, 1995;
Lipovetsky and Tishler, 1999; Zeshui and Cuiping, 1999). However, a number of criticisms have been raised
on AHP over the last several years which are summarized in Table 1. This reveals that the method does bear
some shortcomings in the problem solving domain. Although different variants or adaptations of AHP are
emerged in recent decades, for simplification, this paper tends to focus on its original version which was
developed by Saaty (1980, 1985, 1988, 1990, 1991).
In recent years, construction periods have become shorter and shorter due to high land prices and chang-
ing economic conditions. As a result, substantial liquidated damages have been imposed by developers on
construction contracts to avoid project overruns. Designers and engineers in building and construction are
often required to make speedy decisions to derive an optimal choice from an array of construction solutions
to cope with the changing construction environment and customersÕ needs. Hence, research and develop-
ment of multi-criteria decision-making tools are becoming more important for the industry. In this connec-
tion, this paper attempts to analyze and highlight some weaknesses of AHP in practical applications for
construction. Besides, to exploit more decision-making tools for the current practitioners, an alternative
approach—the Non-Structural Fuzzy Decision Support System (NSFDSS) invented by Chen (1998)—is
compared, aiming at overcoming the weaknesses of AHP. Noteworthy, NSFDSS has been successfully ap-
plied on site layout planning (Tam et al., 2002a). Furthermore, Tam et al. (2002b) have modified the system
to NSFDSS-II for application in construction safety management.

2. Case study

A construction design problem is used for illustration. There is a proposed high-rise commercial building
project located in Hong Kong, 40-storey high with 3 m floor-to-floor height, a site area of 1600 m2, with a
medium degree of complexity of construction. The available options of construction design are steel struc-
ture (D1), reinforced concrete structure using traditional timber formwork (D2), reinforced concrete struc-
ture using proprietary formwork (D3), precast façade systems (D4) and concrete core braced systems (D5).
The decision criteria include time (C1), cost (C2), project complexity (C3), safety (C4), resources (C5), qual-
ity (C6), structural and market considerations (C7&C8).
Using AHP to evaluate the construction design problem, the processes and results are shown in Fig. 1.
The normalized values of priority in original order of decisions are shown as follows:

D1 D2 D3 D4 D5
0.1851 0.1893 0.2519 0.1740 0.1997 (Solution A)
C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324 1319

Fig. 1. Processes and results of AHP in case study.

1320 C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324

Table 2
Comparison of scale of importance between AHP and NSFDSS
AHP NSFDSS
Scale Interpretation Scale Interpretation
1 x and y are equally importance 0.5 x and y are equally importance
3 x has moderate importance than y 1 x is more important than y
5 x has essential or strong importance than y
7 x has very strong importance than y
9 x has extreme importance than y
2, 4, 6, 8 Intermediate values of importance between x and y
Reciprocals of the above x is less important than y at different scales 0 x is less important than y
nonzero numbers

For NSFDSS, its calculation methods and procedures have been detailed in Tam et al. (2002b). Using
the simplified scale of importance as compared to AHP (Table 2), together with the works of Tam et al.
(2002b) to deal with the same construction problem, the output matrices, assignment of priority scores
and calculation of weightings for the problem in question are derived and shown in Fig. 2. Normalizing
the weightings of each decision will give the following values of priority in ordinal order of decisions:

D1 D2 D3 D4 D5
0.1854 0.1975 0.2156 0.1924 0.2091 (Solution B)

Comparing solutions A and B, it can be observed that AHP yields the largest distance (values of priority)
among the various decisions than those of NSFDSS that facilitates decision-makers in choosing the best
option. However, there are some inconsistencies in the values of priority between the two approaches
due to their inherent theoretical and technical differences. This is resulted from the tolerance used in the
‘‘Consistency Ratio’’ (CR) of AHP.

3. Problems of consistency

According to the principles of AHP and previous researches, the global acceptance of the ratio is less
than 0.1, meaning that the result is acceptable if the ratio obtained is between 0.001 and 0.099. However,
such an acceptable range will affect the quality of the solution. For example, considering that there are three
decisions D1, D2 and D3 and their pairwise comparison are D1/D2 = 1/3, D2/D3 = 4 and D1/D3 = 1/2.
The CR in this case is 0.093, still falling within the acceptable range. The problem resulted from this incon-
sistency can be derived using simple mathematical calculations. Suppose the value of D1 = 10, then from
the above relationship, D2 = 30 and D3 = 20. In this sense, the relationship D2/D3 = 4 in the pairwise
comparison is invalid. Two conflicting situations are encountered: either keeping D2 or D3 constant. If
D2 is kept constant (situation 1), D3 = 7.5. On the other hand, if D3 is kept constant (situation 2),
D2 = 80. Consequently, for situation 1, the priority order for decisions is D2 > D1 > D3; where for situ-
ation 2, the priority order will be D2 > D3 > D1. After testing, the result of AHP is same as that for sit-
uation 2. This will be controversial when the actual result is close to situation 1 and the priorities obtained
thereof will then deviate from the optimum solution. Aside from this example, supposing the decision-ma-
ker has made a judgmental error during the evaluation process and the CR of which is 0.046. The corre-
sponding pairwise comparisons for the decisions are D1/D2 = 1, D2/D3 = 3 and D1/D3 = 6. Akin to
the previous example, assume D1 = 10, then the corresponding value of D2 = 10 whereas D3 = 10/6.
C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324 1321

Fig. 2. Processes and results of NSFDSS in case study.

1322 C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324

Similarly, two conflicting situations are encountered. For situation 1, keeping D2 = 10 to be constant, then
the value of D3 will be 10/3. Conversely, for situation 2, keeping D3 = 10/6 to be constant, then the value
of D2 will be 5. Therefore, the priority order for situation 1 is D1 = D2 > D3 while for situation 2 is
D1 > D2 > D3. Again, the result generated from AHP is identical to situation 2. Undoubtedly, the result
is not reliable as the decision-maker has made a wrong evaluation in the pairwise comparison.
In fact, the root of the problem is the 9-point scale of relative importance proposed by Saaty (1980),
assuming that decision-makers should understand well the relationship and the magnitude of differences
of various decisions under consideration. In practice, using such complicated scale is hard to achieve an
absolute consistency in the evaluation process and thus the CR is developed as a guide for decision-makers.
Further, from the current practitionersÕ point of view, another difficulty associated with utilizing the 9-
point scale and CR is the timely process to sort out the inconsistency, if found. It is because the deci-
sion-maker can only know the existence of any inconsistency after inputs are completed and the consistency
ratio is generated. By then, the time spent on completing the whole input table is wasted and the
whole input process needs to be repeated again until the consistency ratio becomes acceptable. Instead,
the use of NSFDSS can avoid this, which provides a more pragmatic approach for the construction
practitioners.

4. Trade-off between merits and deficiencies

1. Automatic consistency correction—NSFDSS facilitates automatic consistency checking. It assumes the

upper rows of the matrix are more reliable then the lower rows and the system will re-set the values
of the lower rows if inconsistencies are found. In other words, the earlier comparisons made are assumed
to be more accurate. The assumption is built upon the belief that if oneÕs attention span is limited, then
earlier judgments ought to be more accurate, but that is not the only condition that will explain variance
in poor judgments. Distractions may occur at any time. The automation process thus may limit the flex-
ibility in the verification process. On the other hand, AHP only provides CR as a guideline for decision-
makers to monitor the consistency of pairwise comparison. Once the global acceptance fails to comply
with, the whole evaluation process needs to be repeated again. This time-consuming approach is unsuit-
able for construction problems where time is of essence in construction projects.
2. Simplified scale of importance—NSFDSS utilizes only ‘‘1’’, ‘‘0.5’’ and ‘‘0’’ for describing the scale of
importance whereas AHP relies on a 9-point scale. The requirement to judge the magnitude of difference
between decisions lengthens the decision process time and complicates the decision procedures. Conse-
quently, difficulties in judgement and comparison cause uncertainty that jeopardizes the accuracy of the
results.
3. Problem of equal weightings—As the scale of importance used in NSFDSS is simplified (with only three
options), the probability of having equal weightings is higher after summing up each row of the output
matrix. In reality, to a certain extent, two decisions may not be exactly identical. Their importance will
be close to each other but are not distinguishable using NSFDSS. As a result, same priority scores will be
assigned.
4. Problem of priority score assignment—As priority scores are derived from the fuzzy set theory, there are
no rules of thumb for decision-makers in judging which score is appropriate to each weighting. One may
find difficulty when assigning a shorter scale of priority scores to voluminous decisions. For example,
Fig. 3 shows that it is hard to assign the scale from ‘‘Same’’ to ‘‘Slightly Different’’ (five numbers of
scores) as determined by the decision-maker to eight numbers of decisions. Hence, the decision-maker
should compromise his behavioral preferences in certain circumstances.
 C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324 1323

Fig. 3. Scenario for the problem of assignment of priority scores.

5. Concluding remarks

The development of decision-aiding tools aims to help decision-makers select best options among dozens
of decisions which sometimes conflict with one another. AHP is commonly used in such analysis but the
drawback of the timely process has been experienced by most decision-makers. NSFDSS is an alternative
to AHP which changes the consistency checking from a ‘‘1–9’’ scale to a ‘‘1–3’’ scale. This change conse-
quently enables the automatic consistency checking process, in which, there is an assumption that the judg-
ment on the upper rows of the matrix is more reliable then the lower. The system will assign the logically
appropriate values of the lower rows if inconsistencies are found. Another change is the addition of a fuzzy
scale of priority scores to obtain the relative degree of difference of each comparison. Practically, the
NSFDSS can save a lot of time in data collection but it may lead to a decrease in the distance of differen-
tiation (the values of priority) between decisions, making it difficult to single out the best option from the
rest. It needs some trade-off between the pros and cons of NSFDSS.

Acknowledgments

The work described in this paper was fully supported by a grant from CityU Project No. 7001213. The
authors also thank the three reviewers for their valuable comments and suggestions to improve and clarify
the meaning of the paper.

References

Belton, V., Gear, T., 1983. On a shortcoming of SaatyÕs method of analytical hierarchy. Omega 11 (3), 228–230.
Belton, V., Gear, T., 1985. The legitimacy of rank reversal—a comment. Omega 13 (3), 143–144.
Chen, S., 1998. Engineering Fuzzy Set Theory and Application. State Security Industry Press, Beijing.
Dyer, J.S., 1990. Remarks on the analytical hierarchy process. Management Science 3, 249–258.
Li, H., Love, P.E.D., 1998. Developing a theory of construction problem solving. Construction Management and Economics 16, 721–
727.
Lipovetsky, S., Tishler, A., 1999. Interval estimation of priorities in the AHP. European Journal of Operational Research 114, 153–
164.
Murphy, C.K., 1993. Limits on the analytic hierarchy process from its consistency index. European Journal of Operational Research
65, 138–139.
1324 C.M. Tam et al. / European Journal of Operational Research 174 (2006) 1317–1324

Paulson, D., Zahir, S., 1995. Consequences of uncertainty in the analytic hierarchy process: A simulation approach. European Journal
of Operational Research 87, 45–56.
Saaty, T.L., 1980. The Analytic Hierarchy Process. McGraw-Hill, New York.
Saaty, T.L., 1985. Decision Making for Leaders. Life Time Learning Publication, California.
Saaty, T.L., 1988. Mathematical Methods of Operations Research. Dover, New York.
Saaty, T.L., 1990. How to make a decision: The analytic hierarchy process. European Journal of Operational Research 48, 9–26.
Saaty, T.L., 1991. Analytical Planning: The Organization of Systems, The analytic hierarchy process series, vol. 4. RWS Publications,
Pittsburgh.
Tam, C.M., Tong, K.L.T., Leung, W.T.A., Chiu, W.C.G., 2002a. Site layout planning using non-structural fuzzy decision support
system. Journal of Construction Engineering and Management 128 (3), 220–231.
Tam, C.M., Tong, K.L.T., Chiu, C.W.G., Fung, W.H.I., 2002b. Non-structural fuzzy decision support system for evaluation of
construction safety management system. International Journal of Project Management 20, 303–313.
Zahir, 1991. Incorporating the uncertainty of decision judgments in the analytic hierarchy process. European Journal of Operational
Research 53, 206–216.
Zeshui, X., Cuiping, W., 1999. A consistency improving method in the analytic hierarchy process. European Journal of Operational
Research 116, 443–449.