Integrating AHP-Fuzzy Model for Assessing Construction Organizations Performance

Transcription

1 International Journal of Architecture, Engineering and Construction Vol 5, No 1, March 2016, 1-12 Integrating AHP-Fuzzy Model for Assessing Construction Organizations Performance Emad Elwakil School of Construction Management, Purdue University, 433 Knoy Hall of Technology, 401 N. Grant Street, West Lafayette, Indiana, USA, Abstract: Organizations performance assessment is a critical aspect in today s project management research. Construction organizations face difficulties in performance assessment, stemming from the uncertain, fragmented, and unique nature of construction industry. Most of the research neglected the different perspectives of construction organizations functional units when assessing their performance. Therefore, the goal of this research is to design a comprehensive performance assessment model through identifying and ranking a set of critical success factors (CSFs). Four assessment models are developed to reflect the different perspectives of four functional units in construction organizations. Analytical Hierarchy Process and Fuzzy Expert System are used for data analysis and models development. The research findings indicate that the CSFs factors in construction organizations have different priorities and weights according to different functional units. The validation results range from 84% to 93%. Overall, performance assessment models will benefit organizations in assessing performance according to the perspectives of different individuals. Keywords: Organizational performance, performance assessment models, organizational functional units, analytical hierarchy process, hierarchical fuzzy expert system DOI: /IJAEC INTRODUCTION Organizational performance is the main driver for success and profit, thus, making performance assessment a necessity to any organization. Moreover, performance assessment of construction organizations is more challenging due to the complex, fragmented nature of construction organizations (Abraham 2002). The factors that affect organizations performance must be fully understood to achieve success (Kaplan and Norton 1995), as well as diversity in the perception of success factors between different functional units within the same organization. A Critical Success Factors (CSFs) evaluation has been found to be the most appropriate methodology to assess and evaluate the organization s performance in order to achieve its main goal of developing a comprehensive monitoring system that contains corporate-wide indicators of success (Holohan 1992; Radwan et al. 2015; Rathore et al. 2015; Radwan and Elwakil 2015). Tsiga et al. (2016) identified 58 success factors that were then classified into 11 groups. These factors were then tested within the space industry using an elicitation technique, using the relative importance index approach to rank the classified categories based on their perceived importance. Babatunde et al. (2016) used the critical success factors (CSFs) to develop a process maturity and determine the current maturity levels of stakeholder organizations in public-private partnership (PPP). The study found that the maturity of CSFs made PPP projects successful. Wibowo and Alfen (2015) identified 30 government-led critical success factors (CSFs) from both the meso and micro levels in public-private partnership (PPP) infrastructure development, measured the importance of these factors, and evaluated the government performance within the Indonesian context. Dang and Le-Hoai (2016) used the critical success factors (CSFs) to identify the correlation between critical success factors (CSFs) and Design-Build projects performance measured by key performance indicators (KPIs). Nilashi et al. (2015) highlighted the importance levels of interdependency among the CSFs which has rarely been explored in the prior studies. In this study, most influential factors in successfully completing construction projects are used to develop a new integrated model, multi-criteria construction projects CSF model. *Corresponding author. eelwakil@purdue.edu 1

2 Many research efforts have been done to determine success factors in construction organizations. However, most of the conducted research only focuses on the construction project level rather than the organizational level (Barakat et al. 2015). Nevertheless, there are various methods and approaches to determine the key success factors of organizations. The most commonly used approach is the utilization of questionnaires and interviews with technical personnel and industry professionals. Overall, the need for determining success factors has increased as it can be an indicator for organizational performance and also can be used to assess and improve performance. The goal of this study is to understand the differences between different functional units perspectives in construction organizations and how the functional units perception affects the construction organizations performance through the following objectives: 1. Identify and study the key success factors for the construction industry at the organizational level; 2. Analyze and determine the weights and impact of critical success factors perceived by functional units on the organizational performance; 3. Build functional units based assessment models for construction organizations performance; 4. Validate the developed models. 2 BACKGROUND Success definitions have evolved over the past decade in the construction industry; and, it is mostly defined as the overall achievement of the organization s goals and expectations. However, success can be assessed differently from one individual to another according to their perspectives. Elwakil et al. (2009) identified eighteen significant success factors for the performance assessment of construction organizations. A regression model based on critical success factors was developed to assess construction organizations performance. The obtained data were analyzed using a back propagation model of artificial neural networks (ANN), which was used to determine the significance of various success factors. Zayed et al. (2012) identified nine critical success factors as the most significant to develop an assessment model for organizational performance. Artificial neural network (ANN) model was used to assess the most significant success factors, as ANN provided the contributing weight of each factor after the completion of the training process. However, there has been a lack of research on the assessment of construction organization performance based on the different functional units perspectives. The available research failed to consider how the different units perceive the success factors differently and thus can affect the performance assessment. Also, the previous research did not consider the integration of different qualitative and quantitative modeling techniques to obtain more firm results. (Rockart 1978) identified critical success factors as critical areas where high performance is important, as these factors decide the success of an organization. In addition, CSFs are the actual steps taken to succeed. Special attention and concern should normally be given to these areas, as these areas can decide the present and the future success of an organization based on its performance (Boynton and Zmud 1984). For the purpose of this research, eighteen critical success factors were identified as the factors that impact construction organizations success. Elwakil et al. (2009), Zayed et al. (2012) classified these factors as the following: 1. Administrative and legal factors include the subfactors: clear vision, mission and goals, competition strategy, organizational structure, political conditions, and number of full time employees; 2. Technical factors include the sub-factors: usage of international aspects, the availability of knowledge, usage of knowledge, business experience (number of years), and product maintenance; 3. Management factors include the sub-factors: employee culture, environment, employee compensation and motivation, applying total quality management, and training; 4. Market and finance factors include the subfactors: quick liquid assets, feedback evaluation, research and development, and market conditions/customer engagement. 3 ANALYTIC HIERARCHY PROCESS (AHP) Analytic hierarchy process (AHP) is a multi-criteria decision making method (Goepel 2013). It is a theory of quantifying intangible factors that affect the decision making process (Zayed and Halpin 2004). It is also a non-complicated technique that attempts to simulate the human decision making process (Saaty 2008). Furthermore, AHP mainly works through a sequence of pairwise comparisons between the factors that influence the decision making process (Al-Barqawi and Zayed, 2008). The of significance AHP stems from its ability to quantify and compare the subjective or qualitative variables (Goepel 2013). In the construction industry, it is very difficult to subjectively evaluate the performance of workers or the effect of certain variations on the organizational performance. Therefore, a need for a method that converts subjective opinions of qualitative numbers is a necessity. AHP has been implemented in many different research fields. Al-Harbi (2001) applied AHP as a decision making tool for project managers. It also has been 2

3 utilized in the selection process of contractors for specific projects based on qualification criteria (i.e. experience, financial stability, quality, resources, and equipment). For instance, Korpela and Tuominen (1996) presented an integrated approach to the site selection process of a warehouse. The study considered both quantitative and qualitative aspects in the selection process. Also, Zhao et al. (2004) applied AHP techniques to simulate and evaluate the relative weighting of the fire safety attributes of buildings. A great feature of AHP is its flexibility to be integrated with different modeling techniques like multiple linear regression, fuzzy logic, artificial neural network, etc. These techniques enable researchers to extract benefits from all of the combined methods, and, hence, achieve the ultimate goal in a more comprehensive way. Three main principles form the basis of solving a problem: 1) developing the hierarchies; 2) setting the priorities; 3) ensuring logical consistency within the factors. To develop an AHP model, however, six steps are required (Al Khalil 2002; Vaidya and Kumar 2006; Saaty 1990; Saaty 2008): 1. Identify the problem to be solved or the purpose of the model; 2. Identify the criteria that influence the behavior of the factors that contribute to problem solving; 3. Assign the relative weights of the factors and subfactors in each category using pairwise comparisons between each pair in the same hierarchy. This requires an (n-1)/2 comparisons, where n is the number of factors with the consideration that diagonal elements are equal to 1 and the other elements will simply be the reciprocals of the earlier comparisons. A comparison matrix is developed as follows: 1 x y Factor Comparison = 1/x 1 z 1/y 1/z 1 where x, y, and z are numbers (integers or nonintegers); 4. Perform calculations for consistency check, if the developed matrix is consistent, then the weight vector for all of the qualitative factors will be calculated by elevating the matrix to different powers and normalizing the matrix (i.e. converting the summation of each column to be one) at these powers. The produced normalized column is the eigenvector. This process is repeated until the eigenvector solutions are not changed from the previous iteration (i.e. up to four decimal places ). If the matrix is not consistent, it has to be returned to the expert to adjust the response and to be consistent with the values. Once it is consistent, step three is repeated. 5. Consistency Index or eigenvalue (CI) is the calculated value used to check the matrix consistency as follows: CI = (λ max d)/(d 1) (1) where λ max is the maximum eigenvector and d is the matrix dimensions. 6. Consistency ratio (CR) is then calculated as follows: CR = CI/RI (2) where CI is the consistency index and RI is the random index, which is the average C.I. of sets of judgments (from a 1 to 9 scale) for randomly generated reciprocal matrices, to indicate whether the estimates are closer to being consistent or to being randomly assigned. According to (Saaty 1990), if the CR is more than 10%, then the results are inconsistent. Thus, the values should be changed until CR is verified; 4 FUZZY EXPERT SYSTEMS Zadeh (1965) introduced fuzzy logic as a powerful modeling technique that can be used to understand the uncertainty of human thinking. Fuzzy techniques have been widely utilized in several research studies over the past decade, and they have the ability to virtually connect humans to computers through analyzing linguistic inputs to stem numerical outputs (Chan et al. 2009). Traditionally, a set of inputs has sharp, crisp boundaries, where elements are either in or out of a set, and the ranking of a membership of a variable is either zero or one (Nguyen 1985). However, in the real world, information is mainly ambiguous and incomplete. Therefore, fuzzy logic comes in hand as elements are allowed to have partial memberships ranging from zero to one (i.e. 0 is no membership and 1 is full membership) (Fayek and Sun 2001). Recently, the fuzzy approach has become popular among construction management researchers as fuzzy logic theory has enabled the handling of complex problems in real world systems, which are mainly defined through linguistic statements. The popularity of fuzzy expert systems is summarized as follows: 1) the knowledge based systems can summarize the human experts experiences; 2) the fuzzy linguistic descriptors are most commonly used by humans, which are inexact and qualitative; 3) these systems can deal with inexact figures and numbers; 4) they provide reasonable decisions even if the input knowledge is incomplete; and, 5) educated assumptions can be used to complete the lack of knowledge in some cases. Chan et al. (2009) categorized the application related to construction management using fuzzy logic research into four categories: decision making, e.g. (Kazaz et al. 2014; Lin and Chen 2004); performance assessment, e.g. (Fayek and Oduba 2005; Zhang et al. 2004); evaluation and assessment, e.g. (Zayed 2005) and modeling, e.g. (Okoroh and Torrance 1999). 3

4 5 INTEGRATING AHP WITH FUZZY EXPERT SYSTEMS Yang and Chen (2004) illustrated that AHP creates and deals with a much undetermined scale of judgment. Furthermore, AHP does not take into account the uncertainty associated with the mapping of human judgment to a number of natural language. Some drawbacks of using AHP solely are that the ranking of the AHP method is rather imprecise and that the subjective judgment of perception, evaluation, and selection based on the preference of decision-makers greatly influence the AHP analysis results. To overcome these problems, several researchers integrate fuzzy logic with AHP to improve the uncertainty. Moreover, numerous input variables complicate the fuzzy model development as it accordingly increases the number of fuzzy rules (Kazaz et al. 2014). If the number of factors is high, the model development would be infeasible as the number of fuzzy rules increases exponentially. In a research study conducted at Purdue University in 1995, Ersoz (1995) integrated AHP and fuzzy logic for productivity estimation; the study introduced a new assessment approach which considered the subjective factors that influence productivity. Another study, conducted by Zayed (2005), integrated the AHP and fuzzy logic methods to develop a productivity index model for piling process. 4. Apply Hierarchical Fuzzy Expert System (HFES) after selecting the most significant CSFs to the different functional units to build the assessment model; 5. Develop a performance scale to assess the organizational performance based on the perspectives of the functional units and how they perceive the CSFs. The AHP is selected because it is a knowledge-based oriented technique that requires the experts opinions to accommodate the success in assessing organizational performance; 6. The models are tested and verified in order to check their robustness in assessing the performance of a construction organization. The different models are validated by applying the models on actual data and by checking the consistency of the output. 6 METHODOLOGY The goal of this research is to develop and validate a model for organizations performance assessment based on the perceived value of the different functional units within the organization. This goal will be accomplished through fulfilling the following objectives. Figure 1 shows a graphical representation of the methodology. 1. Review the existing literature and previously developed models in order to identify the critical success factors in construction organizations; 2. Consult experts to determine the weight of the factors that contribute most to the organizational performance through the opinion of experts in the construction industry. Also, identify the impact of each factor on the overall organization performance; 3. Develop a performance assessment model for the four different functional units in the organizations based on the CSFs using the analytic hierarchy process (AHP) and Hierarchical Fuzzy Expert System (HFES). The AHP will be applied using the data collected from the experts and the pairwise comparisons. AHP will be used to determine the weights of the factors and to select the most significant factors to the four functional units models; Figure 1. Methodology development 7 DATA COLLECTION A number of critical success factors (CSFs) were identified based on experts opinions and experiences. Four main factors were identified as the main categories to be included in the models (i.e. administration and legal, technical, management, and market and finance). Eighteen sub-factors were included in the AHP models, but this number will be reduced in the fuzzy model afterwards based on the significance of the factors to the functional units. The idea is to identify the different perspectives of each functional unit within the organization. Data collection involved two main stages: 1) pairwise comparisons of the main factors and sub-factors 4

5 and 2) identifying the impact of each factor in the performance of the organization. A questionnaire was administered to different functional units in construction organizations to reflect their experience and the company performance from their perspectives. One hundred fifty (150) questionnaires were sent to basic functional units in construction organizations worldwide (i.e. Canada, Egypt, France, Greece, Germany, USA, Saudi Arabia, and United Arab Emirates). The returned surveys from the different respondent groups are shown in Table 1. 8 ORGANIZATIONS PERFORMANCE ASSESSMENT MODEL The AHP model is designed to identify the weights of the CSFs as perceived by the functional units. The significant CSFs are identified for each model and then the HFES model is designed to assess the performance of the construction organizations from the perspectives of four different functional units (i.e. directors, senior engineers, project managers, and cost engineers). 8.1 CSFs Determination (AHP Model) There are several methods to assess the significance of independent factors affecting the performance of a dependent criterion. In this research, AHP method is used to assess the most significant success factors because the AHP method provides the contributing weight of each factor after completing the model training process. Therefore, in the following sub-sections, details regarding these contributing weights are presented. The following steps are a guide for training the AHP model (Al-Barqawi and Zayed 2006). Step 1 - Setting up the factors hierarchy: The factors that affect the organization s performance are divided into three main levels as shown in Figure 2. Level one represents the main objective of the factors (i.e. assessment of organization s performance). Level two represents the four main factors (i.e. administration & legal, technical, management, and market & finance). While level three represents the AHP model, sub-factors, or the overall eighteen critical success factors (e.g. organizational structure, employee culture, environment, and business experience). This step is identical in the four functional units models. Step 2 - Assigning priorities, establishing a priority vector (eigenvector), and response consistency analysis: In this step, the functional units individuals and industry experts provide pairwise comparison matrices for the main factors and sub-factors. Using pairwise comparisons allows the individuals to express the relative importance of one factor over another. AHP analysis is applied to determine the factors weight (W i ) and sub-factors weight (SW ij ) of each factor based on the individuals input. Consistency of the pairwise comparison matrices is tested using Equations 1 and 2 above. The CR values are all less than 10 percent, which is the acceptable range (Saaty 2008). All of the matrices that were received from experts are consistent. This step is repeated for all of the respondents in each of the four assessment models. Table 1. Survey return data Step 3 - Aggregated priority weights: Priority weights aggregation follows the consistency analysis. Where the aggregated weight of each sub-factor is calculated by multiplying the sub-factor weight (SW ij ) by the corresponding main factor weight (W i ) of the same category. Accordingly, priority can be established based on the overall weight using Equation 3 as follows: ASW ij = W i SW ij (3) Where ASW ij is the aggregated weight of the subfactor, (W i ) is the weight of the main factor, and (SW ij ) is the weight of sub-factor j in the ith factor. Table 2 shows the results of the aggregation process based on the average values of the collected matrices of the directors, senior engineers, project managers, and cost engineers functional units models. Step 4 - CSFs selection: The significant CSFs are selected based on their average ASW ij %, as the selected CSFs are equal to or above the average of the total weights of the factors of their corresponding functional unit. Table 2 shows the selected CSFs marked with a bullet. Nine CSFs are identified for each model, except for the senior engineer s model where ten CSFs are found to be above the average. The factors that are not selected are illuminated from the fuzzy model and not considered in the assessment process. It is clear from Table 2 that the administrative and legal category is totally illuminated from the senior engineers model according to the weights of the sub-factors. This proves the robustness of the AHP model, as engineers normally tend to focus more on the other factors rather than administration. Figure 3 shows the relative importance of the subfactors in each model. Where 18 represents the highest ranking and the most important factor to the functional unit, while 1 represents the lowest ranking and the least important factor to the functional units. Functional unit Country of origin Directors Senior engineers Project managers Cost engineers Egypt Canada Other Number of responses Response rate, proportion (%) 19% 32% 33% 16% 49% 33% 18% Total 63 5

6 Figure 2. Hierarchy of the AHP model Table 2. Aggregation process for CSFs Selected CSFs Success factors Directors Senior engineers Project managers Cost engineers ASW ij % ASW ij % ASW ij % ASW ij % X1: Clear Vision, Mission, and Goals 6.28% 4.96% 6.48% 6.54% X2: Competition Strategy 6.14% 4.47% 6.11% 6.02% Administrative X3: Organizational Structure 5.68% 4.61% 5.95% 5.86% & Legal X4: Political Conditions 5.16% 4.10% 5.08% 5.39% X5: Number of Full Time employees 5.01% 3.81% 4.94% 5.02% X6: Usage of International Aspects (ISO) 4.29% 4.36% 4.36% 4.38% X7: Availability of Knowledge 6.3% 5.63% 6.06% 6.06% Technical X8: Usage of IT 5.81% 5.76% 5.86% 5.71% X9: Business Experience (No. of years) 6.43% 6.03% 6.17% 6.43% X10: Product Maintenance 5.02% 5.29% 5.16% 5.38% X11: Employee Culture Environment 5.42% 5.06% 5.29% 5.35% Management X12: Employee Compensation and Motivation 5.74% 6.01% 5.75% 5.76% X13: Applying Total Quality Management 5.25% 5.19% 5.06% 4.97% X14: Technical Training 5.32% 6.16% 5.78% 5.54% X15: Quick Liquid Assets 5.55% 5.71% 5.43% 5.58% Market X16: Feedback Evaluation 5.39% 5.38% 5.49% 5.23% & Finance X17: Research and Development 5.26% 4.71% 4.98% 4.85% X18: Market Conditions/Customer Engagement 5.96% 6.14% 6.04% 5.95% Note: means above average aggregated weights Figure 3. Ranking of the CSFs among the functional units 6

7 8.2 Model Implementation (HFES) The hierarchical fuzzy expert model (HFES) consists of four main sub-models with the exception of the senior engineers model which contains only three submodels. These models are correspondent to the four categories of the critical success factors. Finally, the last sub-model combines the results of the later four sub-models in order to generate the organization performance assessment. The crisp defuzzified results of the four sub-models (i.e. administrative & legal, technical, management, and market & finance) are combined together through the organization performance model. This process is done for the four functional units separately, as the result would be four different performance assessment models. Figure 4 explains the full view of one of the fuzzy models in this study (i.e. directors model). Step 1 - Membership Functions Determination: On the questionnaire, industry experts were asked to select a range of numerical values that corresponded with the linguistic states of both input and output factors in order to construct the initial membership functions. The ranges, shapes, and values of the membership functions for each variable were designed according to the information from the literature, as well as average respondents values. There are many forms of membership functions, such as triangular, trapezoidal, bellcurved, and sigmoidal functions. The factors membership functions are used to convert the crisp input data (e.g. number of employees, years of experience, and usage of technical aspects) into fuzzy numbers. Because this study focused on the qualitative aspects of the factors, only triangular and bell-curved shapes were used, hence most membership functions were accurately represented by those shapes. The factors were evaluated on a 0-10 scale and assigned a number of membership functions (MFs) ranging from five to two MFs. In this study, only a representation of membership functions was presented. Figure 5 shows an illustration of the clear mission, vision, and goals factors membership functions used in this study. Figure 5. Clear mission, vision, and goals membership function Figure 4. Full view of the directors fuzzy model in this study Step 2 - Input Variables: Numerous input variables complicate the fuzzy model development as it accordingly increases the number of fuzzy rules (Kazaz et al. 2014). If all of the eighteen factors were considered in one fuzzy expert system model, the model development would be infeasible as the number of fuzzy rules increases exponentially. If a complete rule base was created for a fuzzy expert system with eighteen input variables and each variable had three membership functions, the number of rules required would be 3 18 (approximately four billion rules). This reasoning supported the use of the AHP technique in order to reduce the number of fuzzy and criteria. As a result, the largest number of rules was 27 rules in the sub-models. Table 3 shows the input variables and their corresponding linguistic variables and the numerical scale used for the directors model. Step 3 - Output Variables: The construction organization performance assessment was the main goal of this research. Moreover, the membership function of the output included five fuzzy linguistic descriptors (i.e. poor, fair, good, very good, and excellent) and the performance was ranked out of 100 to increase the sensibility. The scale shown in Figure 6 displays the scale used to compare the fuzzy numbers. Because the model was developed for performance assessment, the fuzzy sets were dependent, and there was an intersection between the sets and a sequential increase and decrease of them. Hence the center of sums defuzzification method was used in this research. Figure 7 shows the output variable membership function. Figure 6. Proposed performance assessment scale 7

8 Equivalent impact = (fuzzy set weight) of each factor factor weights 10 (4) Figure 7. Organization performance membership function Step 4 - Fuzzy Rules: The last step of the model development is to establish the rules. Fuzzy rules are the conditions of the model. The rules show the correlation between the input and output variables. The number of fuzzy rules in each model is dependent on the number of factors and the number of membership functions (i.e. fuzzy sets). The rules consist of fuzzy prerequisites or antecedents and fuzzy conclusions or consequences. For example, in the directors model: If the Clear Vision, Mission & Goals is very good (7.5); And the competition Strategy is moderate (5); And the organizational structure is very good (9.72); Then the Administrative & Legal aspect of this organization is very good (73.4). The words in bold are a description of the factors as assigned by the experts. The numbers in parentheses are the assigned crisp value (numerical value) from the survey. The sub-model equivalent impact (i.e. fuzzy numbers) shown in the previous example above can be calculated by simply calculating the total number of fuzzy sets multiplied by the weight of each factor then divided by the sum of all factor weights in this sub-model as shown in Equations 4 and 5. It is important to highlight that the differences between the four functional unit models will stem from the different input factors selected from the AHP model. Also, the weights of factors will be different, as each functional unit perceives the factors differently. Organization P erf ormance Assessment M odel = ( ) + ( ) + ( ) ( ) 10 = 73.4 (5) The crisp value from Equation 5 is then compared to the performance scale in Figure 7 to identify the linguistic term of the category. Step 5 - Consequent Aggregation: Similar procedures were established for the rest of the sub-models to finally aggregate the consequences of all of the factors and to generate the final model for assessing the organizational performance. The same concept applied to the four functional units models. However, due to the selection of different factors based on the perception of the different functional units, the models had different performance rules and ultimately different performance models. This process occurred after evaluating each role in the knowledge base and before the defuzzification process. The output linguistic variable was aggregated using a maximum mathematical operation as in Equation 6. During this process, the maximum membership value of any membership function is used to abbreviate the membership function for later ease of use in the defuzzification process (Fares and Zayed 2010). µ R (x 1, x 2, x 3,..., x n, y) = V N j=1 [µ R j(x 1, x 2, x 3,..., x n, y)] (6) where V represents the maximum operation and R represents each of the membership functions of the output (i.e. poor, fair, good, very good, and excellent). This operation is also applied to each of the sub-models (i.e. administrative & legal, management, technical, and market & finance). Step 6 - Fuzzy Models Defuzzification: Converting fuzzy consequents into crisp values can be performed by several defuzzification methods. The center of sum Main Performance Factors Administrative & Legal Technical Table 3. Directors model input variables Sub-Factor Linguistic Scale Numerical Scale Clear Vision, Mission & Goals X1 v. Good, moderate, v. Bad 1-10 Competition Strategy X2 v. Good, moderate, v. Bad 1-10 Organizational Structure X3 v. Good, moderate, v. Bad 1-10 Availability of Knowledge X7 High, moderate, low 1-10 Usage of IT X8 High, low 1-10 Business Experience (no. of years) X9 High, moderate, low 1-10 Management Employee Compensation and Motivation X12 High, moderate, low 1-10 Market & Quick Liquid Assets X15 Good, moderate, bad 1-10 Finance Market Conditions/Customer Engagement X18 High, moderate, low 1-10 Organization Performance Poor, fair, good, very good, excellent

9 method is utilized in this research. In addition, the center of sum method calculates the center of gravity of each abbreviated membership function then averages their weights by their area as a reference as shown in Equation 7. This method is utilized for its ease to program, and it provides reasonable results. Crisp organization perf ormance output = CoA 1 area 1 + CoA 2 area CoA n area n area 1 + area area n (7) where CoA n, geometric center of the area of the scaled membership is a function and area n is the area of the scaled membership function n. 9 MODEL IMPLEMENTATION A sample of construction organization performance assessment models for the directors, senior engineers, project managers, and cost engineers are presented in Table 4-7 respectively. According to the final result of the organizational performance assessment model (OPAM), the performance of the organization can be determined and assessed based on the input from the different functional units. Assessing performance can help organizations identify weaknesses and hence work on developing them. Table 4. Assessment from a director s perspective Factors Sub-factors Administration & Legal X1, 8 X2, 6 X3, 4 Technical X7, 8 X8, 2 X9, 8 Management X12, 8 Market & Finance X15, 4 X18, 4 Organization Performance Good Table 5. Assessment from a senior engineer s perspective Factors Sub-factors Technical X7, 8 X8, 8 X9, 4 X10, 8 Management X12, 8 X13, 10 X14, 8 Market & Finance X15, 8 X16, 8 X18, 8 Organization Performance Very good Table 6. Assessment from a project manager s perspective Factors Sub-factors Administration & Legal X1, 10 X2, 8 X3, 8 Technical X7, 8 X8, 10 X9, 8 Management X12, 10 X14, 8 Market & Finance X18, 8 Organization Performance Very good Table 7. Assessment from a cost engineer s perspective Factors Sub-factors Administration & Legal X1, 8 X2, 6 X3, 4 Technical X7, 8 X8, 2 X9, 8 Management X12, 8 Market & Finance X15, 4 X18, 4 Organization Performance Good 10 VALIDATION OF THE DEVELOPED ORGANIZATIONAL PERFORMANCE MODELS The accuracy of each model will be determined separately based on the output performance value with the actual value determined by the functional unit. From the collected data, 20 percent of the responses from each functional unit is selected randomly and kept aside to be excluded from the model designing process in order to validate the modeled values. All of the models are logically and practically experimented to ensure efficiency in assessing the performance. After building all of the models, the validation dataset is utilized to test the ability of the models to assess the organizational performance. The utilized criterion for the numerical match of an observation is the average percent error (APE) and average accuracy (AC) of a model. If the calculated percentage error is less than or equal to 20%, as this research adopts five MFs for the output variable, each MF represents a range of 20% of the possible values (Fayek and Oduba 2005). The percentage of error will be calculated as shown in Equations 8 and 9: AP E % = [ ] V 1 V (8) (V 1 + V 2)/2 Average Accuracy (AC%) = 100 AP E (9) where APE represents the average percent of error of the model, AC represents the average accuracy percent of the model, V1 is the outcome value, and V2 is the actual value. Table 8 shows a sample of the validation dataset being utilized, as well as the APE and AC for the models. From the table, the accuracy values for the directors, senior engineers, project managers, and cost engineers models are 91.5%, 84%, 90.8%, and 92.5% respectively. These values indicate that the obtained results are satisfactory. When comparing the output validation data from all models, the results show that the cost engineer s model is closer to the actual data than the other models. Figure 8 graphically shows the difference between the actual and modeled values of the overall organization performance for the four developed models, which shows a very close pattern behavior. 9

10 Model Directors Senior Engineers Project Managers Cost Engineers Table 8. Models validation samples Validation Actual Modeled APE APE and AC cases performance performance 26 85(excellent) 72(very good) 17% 25 60(good) 58(good) 2.5% APE = 8.5% & AC = 91.5% 24 85(excellent) 87(excellent) 9% 24 94(excellent) 76(very good) 20% 23 75(very good) 82(excellent) 9% APE = 16% & AC = 84% 22 75(excellent) 78(very good) 3.3% 48 90(excellent) 91(excellent) 1.5% 47 85(excellent) 88(excellent) 3.4% APE = 9.2% & AC = 90.8% 46 70(very good) 75(very good) 7% 52 60(good) 58(good) 2.5% 50 95(excellent) 87(excellent) 8.9% APE = 7.5% & AC = 92.5% 48 80(very good) 74(very good) 8.1% Figure 8. Actual performance vs. modeled performance 11 CONCLUSION Construction organization performance is dependent on several success factors. However, there is a lack of research that focuses on assessing construction organization performance based on the different functional units perspectives and how they perceive the critical success factors differently. The framework for a performance assessment model was proposed to assess construction organizational performance from the point of view of several functional units. Analytic hierarchy process (AHP) was utilized to assess the critical success factors (CSFs) based on the input from the industry experts. In addition, AHP was utilized to assign weights to the CSFs as they were perceived from the functional units before building a hierarchical fuzzy expert system (HFES) to assess the organization performance. Also, HFES was utilized to develop four assessment models for the four functional units based on the different perceptions of the factor weights. The developed models were validated by comparing the output to the actual data regarding the organization performance. The validation of the models had satisfactory results of 93%, 84%, 91%, and 94% for the directors, senior engineers, project managers, and cost engineers, respectively. In addition, the assessment models that were based on the functional units perspectives were compared with each other to identify the differences in the perspectives perceived by the different functional units and to possibly determine the most accurate model for assessing construction organizations performance. As a result, future research will 10

11 apply the developed model and compare the different models developed by different modeling techniques (i.e. ANN) or merge two techniques together to obtain more accurate results (i.e. AHP and regression analysis). The end result of this future research will lead to a new generation of specific and accurate company performance models and fully automated models/systems that might partially replace the expert opinion techniques. REFERENCES Abraham, G. L. (2002). Identification of critical success factors for construction organizations in the Architectural/Engineering/Construction (A/E/C) industry. Ph.D. thesis, Georgia Institute of Technology, Atlanta, USA. Al-Barqawi, H. and Zayed, T. (2006). Condition rating model for underground infrastructure sustainable water mains. Journal of Performance of Constructed Facilities, 20(2), Al-Harbi, K. M. A.-S. (2001). Application of the AHP in project management. International Journal of Project Management, 19(1), Al Khalil, M. I. (2002). Selecting the appropriate project delivery method using AHP. International Journal of Project Management, 20(6), Babatunde, S. O., Perera, S., Zhou, L., Udeaja, C. E., Kumaraswamy, M., and Lam, P. (2016). Stakeholder perceptions on critical success factors for public-private partnership projects in Nigeria. Built Environment Project and Asset Management, 6(1), Barakat, O., Bendou, A., and Martin, J. (2015). Success factors for local development project management case of Taroudant province - Morocco. Journal of North African Research in Business, 2015(2015). Boynton, A. C. and Zmud, R. W. (1984). An assessment of critical success factors. Sloan Management Review, 25(4), Chan, A. P., Chan, D. W., and Yeung, J. F. (2009). Overview of the application of Fuzzy Techniques in construction management research. Journal of Construction Engineering and Management, 135(11), Dang, C. N. and Le-Hoai, L. (2016). Critical success factors for implementation process of design-build projects in Vietnam. Journal of Engineering, Design and Technology, 14(1), Elwakil, E., Ammar, M., Zayed, T., Mahmoud, M., Eweda, A., and Mashhour, I. (2009). Investigation and modeling of critical success factors in construction organizations. Construction Research Congress - Building a Sustainable Future, ASCE, Ersoz, H. (1995). A new approach to productivity estimation: AHP and fuzzy set application. M.S. thesis, Construction Engineering and Management Division, School of Civil Engineering, Purdue University, West Lafayette, Indiana, USA. Fares, H. and Zayed, T. (2010). Hierarchical fuzzy expert system for risk of failure of water mains. Journal of Pipeline Systems Engineering and Practice, 1(1), Fayek, A. R. and Oduba, A. (2005). Predicting industrial construction labor productivity using fuzzy expert systems. Journal of Construction Engineering and Management, 131(8), Fayek, A. R. and Sun, Z. (2001). A fuzzy expert system for design performance prediction and evaluation. Canadian Journal of Civil Engineering, 28(1), Goepel, K. D. (2013). Implementing the analytic hierarchy process as a standard method for multicriteria decision making in corporate enterprises: a new AHP excel template with multiple inputs. Proceedings of the International Symposium on the Analytic Hierarchy Process, Holohan, J. (1992). Use of executive information systems in measuring business performance. Journal of Information Technology, 7(3), Kaplan, R. S. and Norton, D. P. (1995). Putting the balanced scorecard to work. Performance measurement, management, and appraisal sourcebook, 66, Kazaz, A., Er, B., and Ozdemir, B. E. (2014). A fuzzy model to determine construction firm strategies. KSCE Journal of Civil Engineering, 18(7), Korpela, J. and Tuominen, M. (1996). A decision aid in warehouse site selection. International Journal of Production Economics, 45(1), Lin, C.-T. and Chen, Y.-T. (2004). Bid/No-bid decision-making - a fuzzy linguistic approach. International Journal of Project Management, 22(7), Nguyen, V. U. (1985). Tender evaluation by fuzzy sets. Journal of Construction Engineering and Management, 111(3), Nilashi, M., Zakaria, R., Ibrahim, O., Majid, M. Z. A., Zin, R. M., and Farahmand, M. (2015). MCPCM: a DEMATEL-ANP-based multi-criteria decision-making approach to evaluate the critical success factors in construction projects. Arabian Journal for Science and Engineering, 40(2), Okoroh, M. and Torrance, V. (1999). A model for subcontractor selection in refurbishment projects. Construction Management & Economics, 17(3), Radwan, A. and Elwakil, E. (2015). Functional units based model for construction organizations performance. International Journal of Architecture, Engineering and Construction, 4(4),

12 Radwan, A., Rathore, Z., and Elwakil, E. (2015). Functional units based Fuzzy model framework for construction organizations performance. Fuzzy Information Processing Society (NAFIPS) held jointly with th World Conference on Soft Computing (WConSC), 2015 Annual Conference of the North American, IEEE, 1 5. Rathore, Z., Radwan, A., and Elwakil, E. (2015). Framework to identify and evaluate the cause of conflicts within a Matrix organization in construction industry using fuzzy expert system. Fuzzy Information Processing Society (NAFIPS) held jointly with th World Conference on Soft Computing (WConSC), 2015 Annual Conference of the North American, IEEE, 1 6. Rockart, J. F. (1978). Chief executives define their own data needs. Harvard Business Review, 57(2), Saaty, T. L. (1990). How to make a decision: the analytic hierarchy process. European Journal of Operational Research, 48(1), Saaty, T. L. (2008). Decision making with the analytic hierarchy process. International Journal of Services Sciences, 1(1), Tsiga, Z. D., Emes, M., and Smith, A. (2016). Critical success factors for projects in the space sector. The Journal of Modern Project Management, 3(3). Vaidya, O. S. and Kumar, S. (2006). Analytic hierarchy process: An overview of applications. European Journal of Operational Research, 169(1), Wibowo, A. and Alfen, H. W. (2015). Governmentled critical success factors in PPP infrastructure development. Built Environment Project and Asset Management, 5(1), Yang, C.-C. and Chen, B.-S. (2004). Key quality performance evaluation using fuzzy AHP. Journal of the Chinese Institute of Industrial Engineers, 21(6), Zadeh, L. A. (1965). Information and control. Fuzzy Sets, 8(3), Zayed, T., Elwakil, E., and Ammar, M. (2012). A framework for performance assessment of organisations in construction industry. International Journal of Architecture, Engineering and Construction, 1(4), Zayed, T. M. (2005). Productivity and cost assessment for continuous flight auger piles. Journal of Construction Engineering and Management, 131(6), Zayed, T. M. and Halpin, D. W. (2004). Quantitative assessment for piles productivity factors. Journal of Construction Engineering and Management, 130(3), Zhang, H., Li, H., and Tam, C. (2004). Fuzzy discrete-event simulation for modeling uncertain activity duration. Engineering, Construction and Architectural Management, 11(6), Zhao, C., Lo, S., Lu, J., and Fang, Z. (2004). A simulation approach for ranking of fire safety attributes of existing buildings. Fire Safety Journal, 39(7),