International Journal of Construction Management G= ssw: (Pring (online) ournal homepage: wwntandfontine.com/journals/iem20 Regression-based model predicting cost contingencies for road network projects Taher Ammar, Mohamed Abdel-Monem & Karim El-Dash To cite this article: Taher Ammar, Mohamed Abdel-Monem & Karim El-Dash (10 Oct 2024) Regression-based model predicting cost contingencies for road network projects, International Journal of Construction Management, DOI: 10.1080/15623599,2024.2411082 To link to this article: https://doi,org/10,1080/15623599,2024,241 1082 Eh View supplementary material CA EES Pubtished ontine: 10 oct 2024 CB submit your antici to thisjournal BEG view retatedartictes W® view crossmark data Full Terms & Conditions of access and use can be found at https:/Awww,tandfonline.com/action/journalinformation2journalCode=tjcm20psi or 1080/56235692024 2411082 Taylor & Francis cei Regression-based model predicting cost contingencies for road network projects Taher Ammat, Mohamed Abdel-Monem and Karim El-Dash Department of Engineering at Shoubra, Benha University, Egypt AsstRACT Cost contingencies represent a significant markup percentage for construction projects. Many construc: tion projects experience cost overruns due to unexpected events before or during the construction phase. Determining the cost contingency for construction projects is one of the most frequent and wide- Spread financial problems many countries face. Since the amount included in the estimates is always Insufficient to absoro the actual cost overruns of the project. The main objective of this paper is to develop 2 simple realistic regression model for predicting the appropriate contingency amount for road network projects before the bidding stage. The regression model has been developed based on data from real case studies for 75 road network projects to adequately consider all potential risks may face the project. Good-of‘fit tests were used to select the bestfiting statistical distribution for a given dataset Based on the analysis of selected data, the log-logistic probability distribution function is the best-fit curve representing cost overrun behavior, and the cumulative distribution function has been used to determine the realistic probabllty of incurred cost overrun. Based on the analysis of real case studies, it is suggested that an average of 59.722% be used as a cost overrun contingency amount for road network projects. The study outcomes will help practitioners and researchers Improve future accuracy/calculations Of contingency amount for road construction projects and imprave the decision-making process. Abbreviations: AHP: Analytic hierarchy process; 809; bill of quantity; CDF: cumulative distribution func- tion; GARELT: authority for road and bridges and land transport; LL: log-logisic; EGP: Egyptian pounds; IMCS: Monte Carlo simulation; MHUC: Ministry of Housing, Utilities and Urban Communities; POF: probabit ity density function; RA: regression analysis; RBE: risk-based estimation; RNCP: road network construction AnmiCLE HisTORY Received 22 Ap 2024 Accepted 19 September 2024 KEYWORDS Cost overt contingency prediction; regression Fadel, probabiiy stbution, rad networks projects; Rk relative importance index; 3P: three-parameter Introduction Construction cost overruns are an endemic feature of transport tion and road networks. Generally, cost overrun in transports tion projects has heen defined at the difference between the actual and forecasted construction costs (Peter et al, 2015 Segelod, 2018). Cost overrun usually refere to a budget, growth fof cost increase, Therefore, cost overruns are a significant con- cem for the public and private sectors, as they may cause nega tive effects on profits and the economy while also causing considerable discomfort to a large proportion of taxpayers and sharcholders (Love et al. 2015). Cost overrun may be referred to as budget overrun’ or ‘cost growth’ if a cost overrun occurs and the budget is used as the basis for comparison, External factors such as inflation can be predicted from cost escalation and can be used to express the expected increase in budgeted costs (Odeck and Welde, 2021). Cost overruns frequently occur in the construction field because of an inherent risk margin in most projects. Therefore, risk man. agement is an essential process in the construction field (Karlsen and Lereim, 2005) Although many studies have considered the causes of cost overruns, they continue to occur, Design failure, material price ‘variations, inadequate planning, scope changes, and design modi fications are the most significant and frequent factors causing cost overruns affecting road network projects (Herrera et al. 2020), Additionally, other causes, such as not considering uncer- tainties and complexities while preparing budget estimates, espe cially when specific information is available, euch actions are ‘considered unjustified and unfair (Issa et al. 2021) sdrié et al. (2019) concluded that increases in the cost of resources, construction activities, changing design specifications, land acquisition, resetlement, and currency exchange were the key causes of cost overruns in Asian infrastructure projects, Several estimates should be prepared and refined during the ini- tial cost prediction process and before construction. Overall, cost overruns consider the expected increase in construction cost because of changes in market conditions over time while ignor- ing changes in project content, which significantly affect the accuracy of these estimates. ‘A risk assessment process can be used to determine the likel hood of adverse events occurring, as well as the potential impact of these events during a project. Therefore, the estimated impact ‘can be added as a contingency to the project base estimate (Shibani etal, 2022), ‘Agencies usually ignore these models and methods and assume contingency amount as a percentage of the total cost based on intuition and their accumulated experiences. This intut- tive approach of adding a contingency percentage as a contin- {gency amount is significantly inadequate to consider the own« risks in road network construction projects (Baccarini, 2005 CONTACT Taher Arar @ totyaheragrallcom G Deparment of Engineering at Shou, Benha Univers, Eypt @ supplemental data for this article can be accessed online at https:/doi org/10 1080/15623599.2024241 1082. © 2024 loa UE Lite ang Tab Franc GuaeS Creedy et al. 2010), Therefore, owners may sometimes have to take some actions to avoid cost overruns for construction pro} ects, such as ‘© Implementing specific phases and canceling others © Changing or shortening the project path to reduce costs © Modifying certain project specifications or changing standards ‘© Postponing the project until an adequate budget is ‘obtained; and © Canceling the project. Cost contingencies are a substantial markup percentage for construction projects, at many projects encounter cost overruns due to various risks. Determining the appropriate cost conti. gency is a problematic gap and a common financial challenge for many countries, as the estimated amount is often inadequate to cover the actual project overruns Therefore, the main objective of this study is to develop a simple realistic regression model for predicting the appropriate contingency amountipercentage for road network projects before the bidding stage. To predict practical cost contingency percentages, this research analyzed the cost overrun data derived from a survey ‘with experts to identify the most significant factors affecting cost overruns, Based on previous research efforts by authors, the sur ‘vey was conducted to determine the impact of the diferent risk factors on cost overruns in the Egyptian road construction sector (Ammar et al 2022). ‘A regression model was also developed to identify the most critical factors affecting cost overruns for RNCP in Egypt. Best cost overruns estimate and construction contingencies can be predicted by collecting and analyzing enough real data from Himilar previous projects to get the best-fit curve. These data will provide a statistically significant prediction (robust and reliable) of cast overruns for road network construction proj ects, This study vill explore several cost overrun causes with, appropriate analysis to provide a comprehensive understanding of the basic requirements for identifying project risk indicators and provide a reliable predictive model for cost overruns in RNCP, This study also wi answer the following questions: 1, What are the risk factors causing cost overruns for construc tion projects were identified in previous studies that can be assessed before awarding a construction project? 2, What are the common models that have been identified and used by construction experts that have successfully predicted ‘expected project cost overruns? 3, What are the components of the proposed model to predict ‘cost overrun and contingency percentages for road network ‘construction projects across range of projects based on data available before contract award? 4, What is the predictive accuracy of the proposed model? The study contributes to praclitioners and researchers to improve the accuracy of prospective calculations for contingency amounts in road construction projects and to enbance the deci Hon-making process. The study is limited to Egyptian road net work projects. The data was collected from projects in Egypt. Therefore, these results may not reflect all the characteristics of such projects in other countries. road_network Literature review Cost overruns ‘Cost overruns are the amount in which the projec’s actual cost ‘exceeds its estimated costs, with constant prices, and against a fixed baseline (Flyvbjerg et al. 2018). The presence of uncertainty in the construction field is the primary reason behind cost over- rans. The cost overrun percentage can be calculated by Equation (1) as follows (Andrié etal, 2015); Cost overran % ~ [(Actal costs costs Intl cost] 100 o Cost overrun may occur due to ‘internal factor, such as the ‘estimation method, design, project location, complexity, size, and duration, or ‘external factors, such as taxes, regulations, and inflation (Yousry Akal 2017), Nearly 90% of the transportation projects faced high completion costs compared to thei original budget, leading to cost overruns (Love etal. 2015) Contingency Contingency is a risk management tool to reduce the impact on project cost if a certain risk occurs under expected conditions (ALShaalan, 2020; E-Kholy et al. 2020). The contingency allow- ance will vary according to the project (Jackson, 2012). Typically, the estimated costs of the project are determined according to the basic estimates plus the contingency amount, while the base ‘estimates are calculated during the conceptual design phase of the project. From the oviner's perspective, contingencies in con- struction projects can be defined as ‘the budget set aside to deal swith uncertainties during construction, Ginhan and Arditi (2007) stated that an owner's contingency is an additional amount to be included in the owner's project budget, considering the riske and uncertainties that could increase the project’ costs. For example, the estimators include a design contingency for the expected increase in the project costs during the pre-construction phase; meanwhile, the contractor's ‘contingency is included in the construction project budget to ‘overcome unexpected events that the project may encounter dur ing the construction stages. They proposed a methodology that allows owners to systematically take advantage of accumulated project data related to similar projects to reduce unwanted events during construction, This involves taking necessary measures at the pre-construction stage. They conducted a ease study to ana lyze an owner's contingency budgets and actual spending in nine parking lot projects, They found that the systematic approach proposed in their study will minimize the owner's contingency budget From the owner's perspective, cost contingency is an allocated amount to overcome uncertainties and risks that may lead to ‘cost overrun, Therefore, contingencies can be interpreted a the amount of money added to the basic budget considering events that are difficult to identify at the early stages of the project life ‘gee (Tseng et al. 2008) Estimating the risk contingency Usually, a project's estimated costs consist of the basic estimate plus the contingencies, As shown in Table 1, baseline estimates constitute a large percentage during the project's conceptual design phases, which are gradually reduced during the following phases and to the lowest possible limit during the project2005 ove eta, 2015; Kaen and Lere Wel-known projects levels and cost certainty fer construct Table 1, Recommended contingency, q 5| s Recommended cantingency () cont Level of Contact phase (Gatement of se, function bret, scheme Sign, ay dy, detaled design, B00, tender auction and Pro Design contingency rawings, commissioning. ané handover Very hgh Very ow (Operation and handover e implementation phase. At the beginning of a projec, the contin- gency allowance is relatively large owing to more uncertainties and is gradually reduced to adapt to costs incurred during pro- ject phases, ‘As shown in Table 1 (Love et al. 2015), two fundamental types of contingencies exist in any construction project 1. Design contingencies assigned for design changes and modifi- ‘ations, considering some factors, such as data, lack of dar- ity of the scope, and inaccurate estimation methods (Smith et al. 2014), As the project becomes more specific, design contingencies are accommodated in the budget for certain cost items, causing changes in the project's base estimates GBaccarini and Love 2014), Giinhan and Arditi (2007) stated that a contingency of 83% and 40% should be allowed for inaccurate estimates and an incomplete scope, respectively, during the initial budget stage. 2. Construction contingencies consider changes during the con- struction process wherein unresolved design issues befor awarding the contract are incorporated into the estimated project cost (Love and Tka 2021). The construction contin gency exists (o salily these permissible variations according to the contract between the contractor and the project’: Previous studies have confirmed that from the contract awarding point, design modifications, changes in the projects scope, and rework ean contribute to an additional 20%-75% of a project's initial contract value (Ayub et al. 2019; Love and Tka 2021). Generally, a total contingency of approximately 55% for ‘well-known and mature projects should be added to the initial Dudget estimate, and approximately 90% should be added for projects with significant uncertainty from the feasibiity to execu- tion stages (Karlsen and Lereim 2005), as illustrated in Table 1 Table 1 illustrates the recommended contingency percentages based on uncertainty, risk levels, and cost certainty for any con: struction project from the ideation to handover stages. 24, Contingency estimation methods Determining appropriate contingency costs requires enough understanding of how estimators determine budget contingen- cies. Contingency costs are expressed as a percentage markup of the base estimate and are usualy used to consider the potential occurrence of unexpected events ‘Many estimation techniques and methods have been used to calculate cost contingencies for construction projects (Hammad et al. 2016; Dykman et al. 2019). Dykman et al. (2019) identified that the traditional fixed percentage approach is insufficient, They suggest that this could be one of the main reasons why construction projects often go over budget. However, it is unclear which method is the most reliable. Their study provides a focus for future research on the reliability and utilization of contingency methods in the construction industry. Hammad et al, (2016) introduced new measures in the proposed cost con. Lingency management method to control expenditures, balance activites’ cost contingency, and estimate future needs. The new method was applied to a reallife bridge maintenance project. The results show that the proposed methodology is adaptive and robust enough to accommodate typical disruptions in the project environment2 Accurate cost estimates are critical for decision-makers, espe cially in large construction projects, but unfortunately, cost fore casts are often inaccurate (Cantarelli and Flyvbjerg 2013), Previous studies such as (Hammad et al. 2016; Dykman et al. 2019) provided several solutions to mitigate the risk of inaccur ate cost forecasts as follows: ‘© Improving accountability in infrastructure investments and focusing on the following measures: explication of regula tions, involvement of risk capital, performance specifica tions, and transparency. ‘© Improving prediction methods based on decision-making theories involving uncertainty and concerns according to the reference class forecasting method, The accuracy of this pro cess may be ensured by basing forecasts on the performan: cet of comparable, previously conducted projects Contingency estimation and risk analysis in the construction industry ‘A detailed quantitative assessment of risk is usually known as risk analysis. This process involves specific and meaningful levels of potential risks and then quantifying the impact of those risks i terms of quality, time, and cost. The risk analysis process se an essential component in the cost estimation process. All con- struction projects incur risks from inception to the delivery and implementation phases. Some previous studies have considered risk analysis techni ques that can be applied in the construction field, including the following: ‘© Claudius et al. (2015) conducted a study to investigate using ‘Monte Carlo simulations (MCSs) for cost estimation in con- struction projects. They analyzed construction project cost estimates using MCSs and investigated the probability distr: bution mode for diferent cost elements in various construc tion projects. They found that historical data for MCSs ‘could be used to estimate the variation in costs, which could bbe helpful for risk managers and estimating construction project costs. Furthermore, their research showed a positive skew in the cost distributions and several interdependent relationships among the cost elements © ELTouny et al. (2014) used analytic hierarchy processes to identify the financial implications of risk factors that affect ‘contingency costs during the tendering stages of Egyptian highway construction projects. They developed a reliable ‘model to estimate the expected contingency costs, They tested the model using real data from the completed high- ‘way construction projects, which showed that 96.31% of the average estimated contingency for the real projects matched the expected contingency cost. Therefore, their results dem: “onstrated the robustness of the developed model. ‘© Liu and Zigrid (2010) used the risk-based estimation (RBE) technique to determine the accuracy of cost estimates for Australian infrastructure projects. A comparison of several ‘onstruction projects using traditional estimation techniques with RBE for water-based projects showed that contingency ‘estimation for projects using RBE had better accuracy than ‘estimation based on traditional methods. Additionally, pro fects using RBE ate likely to be under budget, whereas pra} fects using the (aditional approach are more likely to be over budget. 3. Research method 5 overruns are an endemic feature of road construction pro} fects (Love and Ika 2021). Construction contingencies generally ‘cover the increase in costs that may occur during the construc- tion stage. To successfully complete construction projects and reduce the tisk of increasing contingency, having accurate pre- dictionsicalculations of cost overruns is a must "To improve the accuracy of the contingency estimations of construction projects, the empirical distribution of cost overruns should be examined to determine the most appropriate probabil ity 20 that the proper amount of contingency can be calculated GBaccarini and Love 2014). The contingency accuracy can be measured by comparing the actual final cost against the pre- dicted final cost (Baccarini and Love, 2014). Based on a specified confidence level, the owner can select a contingency cost to over ‘The research objectives will be achieved through the five: step research procedures, & shown in Figure 1: © Literature review: Perform an in-depth analysis of previous studies to identify factors affecting cost overruns for road network projects in both developing and developed coun- tries worldwide # Data collection: Collect data through questionnaires, govern- ment sectors’ databases, and some case studies to identify the most influencing cost overrun factors in Egyptian NCP, Probability distribution type selection: Choose the best-fit probability distribution type to describe the behavior of cost # Date analysis: Perform descriptive analysis and analysis of variance using SPSS software and discuss the analyzed data © Model development: Develop an accurate, simple, robust, and reliable model to predict cost contingency percentages for road network projects in Egypt Data collection For any construction project, cost performance is a key success criterion for the project owners and can be measured by com- paring the final project cost to the project budget. Therefore, ‘cost contingency estimation is a crucial aspect of any construc lion project. Several techniques have been used to collect and Reviewing literanae abot cost overran contingency ——_—__ 4 Cllecting dats wig qesiouates, soveraneaT socom databases and ‘te ses aod network prose —______ ‘Chocsing he best potty Gtibtoa wo dsb i balnior fom i Abalying and Goes ds, and suggesting he equed cst ings percentage VV Developing the modelo pest cost outingeeyvales Figure 1. Research proceeuresvalidate cost overruns, including questionnaires, real case studies, and governmental databases (Flyvbjerg et al. 2018; Odeck, 2018). Two data sources were used in this study: © A questionnaire survey was adopted from (Ammar et al, 2022). The construction experts’ respondents represented different stakeholders in Egypt's road network construc tion field. The heterogeneity of respondents was achieved by selecting respondents with different roles in the road construction industry, The questionnaire war developed based on two types of data, The first type consisted of, well-documented and peer-reviewed cost overrun factors available in previous literature. The second type included significant data from existing road network projects, such. as project location, size, duration, and detailed informa: tion about the root causes of road construction cost over: runs. Various methods, including face-to-face meetings, telephone discussions, and email, were utilized to gather data from experts Regarding research ethics, al participants voluntarily agreed to participate in this study and provided all data requested in the questionnaire form, © Data collected from Authority for Roads and Bridges and Land ‘Transport (GARBLT 2023), MHUC, and real feld supervision data of completed 75 road network projects such as highways, ring roads, toll roads, and primary inter- city road projects, including bridges and tunnels (Ammar ct al, 2022; World Bank Group, 2018). The collected data include drawings, Bills of Quantities (BOQs), and final statements (ut of 96 finished road network projects (GARBLT), 75 real projects (78%) were selected for data collection, analysis, and val idation for this research, To develop the best-fit statistical distribution, real cost over: run data were used to identify the probability of occurrence before the bidding stage. Comparing the actual final project cost and predicted final cost overrun checks/verfies the accuracy of the proposed prediction model, A smaller difference between these two costs will indicate/measure the accuracy of the pro: posed model. Therefore, the proposed regression model will be used for verfication/validation. “The selected criteria were based on size, duration, and project location to reflect Egypt's road networks. The selected projects include those in the Upper and North Egyptian regions. The actual cost overruns of these projects were determined based on real data from the estimated (project price) and actual costs of these projects Probability distribution ‘The probability distribution represents the probability of consid ering an uncertain quantity of any value from a range of values. Choosing a specific distribution shape to represent the uncertainty assessment regarding a part ofthe project cast is not subject to any objective basis. Most cost estimation accuracy ranges are non-symmetric; therefore, a non-asymmetric distribu: tion is mote realistic for many cost estimates (Markiz and Jrade, 2022) Distribution types Normal distribution is a continuous probability distribution ofa random variable with a real value, tis also known as a Laplace-Gauss distribution Log-normal distribution is a continuous probability distribu- tion of a positive random variable in which the logarithm is normally distributed Logistic distribution is a continuous probability distribution. I has heavier tails than the normal distribution (Figure 23) Weibull distribution is a continuous probability distribution that fits sa extensive range of distribution shapes. It can be used in reliability as a model for time-to-falure Log Logistic (LL) distribution (three parameters [3P]), which fs alto referred to as the ‘Fisk alstribution.’ In economic applications, it can be defined as 2 probability distribution fof a non-negative random variable that includes the loga- rithm of the logistic distribution, The relationship between the LL and logistic distributions ie the same as that between the log-normal and normal distributions (Figures 2(b and 6) ‘The logarithmic distribution has many distinguished properties compared to other parametric distributions, such as survival and reliability (Muse, et al. 2021) ‘The tails of the LL distribution are heavier than those of lognormal distribution, where the tal properties depend on the inference. In contrast to the log-normal distribution, the LL cumulative distribution function (CDF) has a closed form, which makes it more advantageous over the log-normal distribution when handling censored data (Figures 2(d and). LL has @ non-monotonic hazard function, where the hazard function is unimodal when (shape parameter >1) and then ‘monotonically decreases when (shape parameter <1); which makes LL distribution distinct from the Weibull distribution (Figures 2(F and g)) ‘The LL disteibution is a reliable and vital model as it fits well in many practical silvations when analyzing data reliability, ‘Therefore, the probability density function (PDF) of LL. was the bes 2021). The data used in Figure in describing the behavior of cost overruns (Muse, et al. are from cost overruns for the 75 real road network construction projects, Best-fit distribution ‘The ‘bestft’ probability distribution involves selecting the ‘best- fi’ statistical distribution for dataset resulting from a random process. Using EasyFit 5.6 software and the goodness-of-fit, the “best-fit” distribution was determined for the 75 sample projects data a8 a reference for the probability distribution. 3P distribution statistics are used to describe the goodness-of- fit as follows: 27. Chicsquared test, which is used to examine whether a data sample matches a population with a specific dlstsbution. Z: Kolmogorov-Smirnov test or D test that is based on the largest vertical difference between the experimental and the- retical CDF. I is a more powerful alternative to the X° test when its assumptions are met A*: Anderson-Darling testis used to compare the observed CDF data with the expected values. I is considered one of6 © Tawnwer aL 4a, Normal and logistic PDF ‘¢. Normal and lognormal PDF « Log-logistc and log-notmal CDF. Logistic and Log-logistic PDF. Note. 3P: three-parameter 4. Log-logistic and lognormal £Log-logistic and Weibull PDF. Note. 3P: three-parameter 8, Log-logistic and Weibull CDF. ‘Note. 3P: three-parameter Figure 2 2 Normal and legitic POF. b. Logistic and Logdogist POF. Nat. 3°: thee parameter. Neemal and lognanmal POF. é. Lagsegiste and lagnrmal POF «elog logistic and lognormal CD. 2¢ og-egstic and Well POF. § Loglegsic and Webull CF. foe 3 epee the most powerful statistical tools to detect most departures from normality and provides more weight for the tals of the distribution compared to the Z-test. A smaller value of AP indicates that a distribution fils the data better (Gurendran and Tota-Maharaj, 2015). Data analysis and ranking Fifty-six cost overrun risk factors were identified and divided into internal and external groups (Ammar et al. 2022). Typical probabilty-impact matrixes were used to represent internal andwml] Zl a jor} 4 ® [es] = ew) oal a INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT @) 7 Risks Figure 3. frame of» probablty-impact mata with a scoring scheme. Impact reg Dopgistic GP) ‘Kolmogerov-Smirnov Sample Size 5 @ 0.08 01 Critical Value os 0.19 Reject? Anderson-Darling Sample § 75 @ 02 On 0.02 08 Critical Value 14 19 33 Reject? No No No Chi-Squared « 02 on 0.05 | 0.02 0.01 Critical Value 86 1 13 15 7 Reject? No Ne Ne Ne Ne Figure 4 Sodrev at ras external risk factors. Figure 3 shows an example of a probability. Table 2. Top 10 eos overun rik factors (Ammar etal 2022, impact matrix with a scoring scheme. The severity for each risk faking factor can be calculated based on Equation (2) (Rodi 2023). Facts eusing cost overun tank isk Srey = (Llelood) Probability (P)+ Impact 2) @) FR Paarl eine 7 ‘The probability impact score was calculated for cach risk to FS. Quenly erangs eras) 3 determine the relative priority of each risk factor. Risk severity FI, Varition ore 4 ‘was then calculated to obtain the severity degree for each risk fp peye‘enes i factor. RII was used to prioritize all risk factors (Khatib et al. F17 Speciation changes 7 2020; Morad 2023), as shown in Equation (3), and factors were IB. Scope of work : then ranked based on their RIT values 9 Unreal ontion : a= > w/(AeN) o Where: W, is the weight given by the respondents for each factor (ranging from 1 to 5) A is the highest weight (5 in this study) Ni is the total number of respondents (75 in this study) Table 2 shows the top 10 ranked factors causing cost overrun for Egyptian road networks, RA is commonly used to predict construction costs It is espe- cially effective for the early determination of cost estimates when project details are rare, It can be used to estimate the project's ‘cost contingency, which is the key component of the overall pro ject budget (Baccarini and Love 2014). Regression models out perform other prediction models from a practical point of view ‘when the actual contingency cost is compared to the expected values (Baccarini 2004), ‘Zurada et al. (2011) conducted a study comparing regression and artificial intelligence methods, specifically neural networks (NN), radial basis function neural networks (RBENN), and ‘memory-based reasoning (MBR) in a Mass Appraisal Context ‘The results showed that regression-baced methods outperformed2 artificial intelligence methods im all simulation scenarios, partic Tasly with homogeneous datasets Yildz et al. (2017) implemented and compared regression and other machine-lesrning models for commercial building clectiity load forecasting. They found that the regression mod dls performed fairly well in comparison to other more advanced ‘machine learning model |-Kholy (2015); E-Kholy et al. (2020) and Abu ELMsaty ct al (2017) conducted three studies comparing the accuracy of regression techniques with other approaches © Im 2015, they discovered that the regression model had higher prediction capabilities than the case-based reasoning (CBR) mode! ‘© In 2017, they recommended using linear regression-based models over statistical fuzzy-based models for predicting cost escalation in highway construction projects in Egypt. © In 2020, they concluded that regression-based models are ‘mote practical than ANN models. Therefore, contractors are advised to use regrestion-based models as they outperform [ANN models in estimating cost contingency. The coefficient of determination R° value is a good measure of the objective magnitude of the relationship between the dependent and the independent variables (Field 2020). Ris a measure used in statistical analysis to evaluate how well models, explain and predict future outcomes (Cheng et al. 2014). The coefficient of determination R’ is more truthful and informative than the symmetric “Mean Absolute Percentage’ (SMAPE). does not have the interpretability limitations of ‘Mean Square Error’ (MSE), ‘Root Variant of Mean Square Error (RMSE), Mean Absolute Error (MAE), and Percentage Variant of Mean Absolute Error (MAPE). Therefore, Ris preferred as a standard metric to evaluate RA in any scientific field (Chicco etal. 2021) In regression models, there are two types. of variables: dependent and independent and the relationship between them. can be represented graphically. The regression equation is used. to predict the values of the independent variable. The depende variable is the outcome variable, while the independent variable is the predictor variable. The regression equation is a statistical, model to determine the specific relationship between the pre dictor and outcome variables. The regression equation model allows for predicting the outcome without or with minor errors The regression equation contains numerical relationships between the predictor and outcome variables “The regression model has been used to measure the overall impact in cost overruns due to individual factors, as shown in Equation (4) (Baccarini and Love, 2014). VSB BX bo Bike $C ® Where (¥) is the dependent variable: (B;, Bp, and B,) are regression estimates; (Xi, Xa, and X) are independent variables; and (C) isa regression constant The optimal model is selected based on the strength of which i a direct measure of variance Results and discussion All analyzed project data were collected from the public sector ‘and represented most Egyptian governorates (Upper and Northern Egypt). Cost overrun data were analyzed for 75 road petwork construction projects, and a unit-price contract was used in these projects. Data from 75 actual projects were used to develop the model. Data from 12 of these projects have been used for validation and selected randomly to test and validate the proposed model Based on data analysis, the minimum and maximum cost ‘overruns were 18,18% and 181,02%, respectively. The mean cost overrun and standard deviation were 59.722 and 25.238 million, LE, respectively. Using SPSS 25 software, descriptive analysis was performed using the basic data collected from the survey. Analysis of variance was used to examine whether there was 4 difference in the cost overrun between projects with different price values. No statistically significant differences were found between the project cost and cost overruns [F (2, 74) = 0.416, p< 005} ‘Table 3 summarizes the descriptive statistics at a 0.05 signifi- cance level for the sampled projects. All 75 projects cost more ‘than 100 million Egyptian pounds (EGP), wherein the estimated values of the studied projects ranged from 103 millon to 2.418, billion EGP, as presented in Table 4 EasyFit 5.6 software allows the possibilty of fitting « large ‘number of distributions automaticaly or manually and selecting the best model Using the EasyFit 5.6 software, the goodness-of-fit test results that utilized the Hy value (wherein the datasets followed a spe- cific distribution and were not rejected) for sample distributions (Ge. the best fit) at 0.2, 0.1, 005, 0.02, and 0.01 a levels are pre sented in Figure 4, The ‘best fit” probability distribution was ‘examined using X°, Z, and A” tests, and results of the goodness- ‘of-fit tests clarified that the LL-3P method presented the ‘best-fi for the sample projects at a < 0.05, as ustrated in Table 5 Figures 4 and 5 show PDF and CDF for the LL-3P for the considered 75 road network projects, respectively Probabilities derived fom statistical distributions are used as the basis to assess the sufficiency of construction contingency ‘costs. For example, using the LL probability of a cost contin gency af 10% would be sulficient for being considered in the cost overrun calculations, which was found to be 35.39% as shown in Table 6. In this case, there is a 64.61% chance that the project will experience a cost overrun if a 10% cost contingency is included. Conversely, based on a specified confidence level the owner may want to establish a cost contingency for ths risk such as when the owner wants to set a contingency cost of 50% probability (Psa) to sulicently consider the potential cost over run, In such cases, the LL system indicates that 56.88% contin gency costs is required. Whereas in the case of Pho, @ cost contingency of 75% would need to be allocated, while for Poo ‘ease, a cost contingency of 84.03% would need to be allocated. ‘Therefore, the intuitive approach of adding a contingency per centage (10%) will not be sficient to cover the owner's risks. For public sector agencies, adding an 84.03% cost contingency to a fixed-price contract may seem excessive, but it will cover the errors and omissions that will arise in the contract docu- ‘ments and the rework required during the construction phases Based on the above analysis, using a construction contingency Table 3. Descripive statisti for cost overuns. Stati Value Stati Vale Sample sae 75 om 24198 Mean som. 10% 45398 ange rote 2% an Medan (50%) seas 158 oar Sandia deviation 2578 oom son Variance 63638 os 166 Kener 7307 inimum Tee ‘skewness 1906 Maxum werINTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT @) 9 Table 4. Actual ost overrun percentages forthe sample road nexwork projets. Project Estimated costs (Price) lion LE) ‘Actual costs (ion LE) ‘ost overun (8) mr 7387 5180 3636 a 135570, sso asi cy "0000 "1200 aa "5 220818 301500 3658 % 150.00 245.00 an ” snes? $8280 a8 me 22400, 521.00 2026 7 45000, 60225 nis Pio mas 27s man my jan0 43200 a0 pz 10500 300.00 502 Pa 830 270000 5532 re 700 S180 seas ms Wa30 2sa0 eat 6 28000 77450 a7 Pr 18.00 22900 3631 rie ss100 S430 ase m0 12960 20180 571 mI 1200 22500 mas ma 4200 25430 ea36 mB. 157730 21600 no ne 27500, 453.00 an ms 9737 30450 sae m6 100 sr300 sass mr 126818 3.40 BD on 200, 351.00 557 i) Sago a0 5139 Po auaao 56570 da a 241800 as4400 bei Pa 36000, 557.00 sana Pa ‘rsa0 te200 na ms 58500, 57500 435 76 93700, aa300 47 mr 26800, 400.00 sage me 195680, 2683.00 um i) 12400 21150 7056 P40 23500 Siaso se Pat 135600, 192200 a8 par 25600, 40200 5703 Pa asra0 22350 or Paw ‘e200 546.00 ais as wae 235 a7 a6 25700, 22560 6560 par 208s 37940 S12 Pas Wo00 seaco sen 50 3si00 55050 7073 51 12900 20198 ses? Ps 18200 35480 276 58 24500 35450 40 55 152431 148596 239 6 21800 392.00, na "57 16785 31100 8525 58 4300, sso 5.00 59 43400 20980 seer eo 36180 sen 20 61 24000 30200 230 rer mma ssox07 32 Pea 102700 365.00 Bn 765 31530, ana 3974 P66 29126 a0 rae er 00 42100 fan 768 10030 19450 7359 770 405.10 158.04 sa Pn eso 48200 30 mm 13628 21840 sas mm 26400 42230 5396 re joa70 18530 yaar mS 2s ass 48610 © TAMAR ETAL value higher than the average cost overrun of 56.88% is recom- mended to account for known unknowns such a& omissions and errors in the contract documents. From the owner's perspective, design errors and omissions are identifiable risks that should be included in a construction contingency. ‘The probabilities obtained from satitical distributions serve as the basis for mon- oring and re-evaluating project costs, Road agencies can imple ‘ment appropriate risk management strategies by identifying the key factors that contribute to cost overruns from the contract award stage and assessing thei likelihood of occurring. Table 6 presente discrete probabilities extracted from the CDF, The probability that road network projects will encounter an average cost overrun of 59.7228, as a percentage of the actual “Cont over 6) fr Road Netwonk 1 Figure 5. The POF af cost ovens forthe rod network projets Comsinive Dstibtion Fonction Prottily . ‘Cost overrun (%) for Road Network Projects Figure 6. The CDF of cost overruns for the road network projets Table 5. Goodnesoft tet for 46 networks project cost, is less than 57%. Typically, estimators have no stat- istical knowledge of the suitability of a deterministic percentage approach for cost overrun when using such an approach to cal- culate the cost contingency. Compared with previous literature, Ayub et al. (2019) and Love and Ika (2021) have stated that an additional 20% to 75% ‘of a project's initial contract value is necessary to consider scope changes, design modifications, and reworks, Additionally, Karlsen and Lereim (2005) found that a total contingency of 55% should be added to initial budget estimates for mature and ‘well-known projects, while a 90% contingency should be added to projects with significant uncertainty from the feasibility to the implementation phase Based on this analysis, to adequately consider all potential risks, a contingency value above the average cost overrun of 459,722% for road network projects in Egypt is recommended. 4.1, Develop a simple regression model RA is an estimation approach used to determine the contingency ‘ost for the project, which is an essential component in project budgeting (Baccarini, 2004), RA is commonly used as a predict- ive approach to examine the bestftting parameters ftom avail able system data points (Alzobky 2022) ‘A forward step-wise process of SPSS 25 software is used to formulate the regression model, which may not always result in, the best model, Using Spearman correlation, predictor variables that show a significant association with the dependent variable are identified. Various models are then reformulated using fo ward and backward step-wise entry methods. Since the value of changes rapidly with the addition of new independent vara bles, a good measure of model quality is the adjusted R? value Ideally, the R? and adjusted R? values should be equal (Arkes 2022). The difference between R? and adjusted R? represents the predictive power of the model. A smaller difference between the {wo values indicates a more robust model (Feld 2020), "The optimal model was selected from the resulting models based on the strength of the coefficient of determination 3 value, which is a direct measure of the variance (Elfabham 2018; Field 2020) ‘The displayed values in Table 7 are acceptable with reasonable strength, depending on the goodness-of-fit of the model, where +The model can explain 9896 ofthe cost estimation using the identified variables based on the strength of the value + Because the value of R changes rapidly as new independent variables are added, the value of the (adjusted R) is a good, measure of model quality (Field 2020) ‘+The difference between the (adjusted RP) and RE represents the predicive power of the model, where a smaller diffe. cence between the two values indicate a stronger model (ield 2020) Distibiton ype Semple sie « ‘Chisquare F)___Kelmogorev-Smimev i Anderson-Daring OF) Log-Logisie 87 75 oor 1s085 017268 roy oor ena bese aoa Table 6. Discrete posses of cost overruns for Eaypian RNCP. Probably 8) 5 10 » 30 7 o 7 w 0 Costovenvn () DISS SSBB STSTable 7. Regression model cefciens Vaile (B)Coetient Signeance @) eae intercep oon ° 45 Poor qualty material supply 0235, ° aseoss Fa Seope of work onsi ° 1S Quantity changes (rene) ors ° #3 Poor quay of work 016 001 27 Delay dept resotion ov ° 28 Inproper version of contract documents oxs7 002 1S nacurate cost estimate 0081 ° FI Lack of coordination 101 oon Table 8. Probably of contingency casts based on the regression model cue Tnercest FS Fie AS 7 7 7 7 aa Cost ovetun % 7 3 22 var a 16 7 187 3 103 1s 3 3 o ° 3 16 7 157 3 1, 65 4 3 6 ° ° 16 "7 187 3 103 62 5 a 6 ° ° 3 "7 187 a 103 82 7 3 ° ° ° ° 0 ° 3 10, 205 ‘ 3 6 Q ° 6 6 ° e 103 na © The regression coefficient of all factors is strongly significant with p< 0.05 (Field 2020). ‘© Standardized beta coefficient fb compares the strength of the ‘effect of each individual independent variable to the depend: cent variable, as shown in Table 7. The independent variables (predictors) with a higher B value are more important to the regression model than those with lower values. The final regression model for determining the cost overrun is presented as shown in Equation (5) Predicted cost overrun — 0.295 P45 + 0.181 F18 40.173 PIS $0,160 F43 4 0.170 F27 40.157 Fe $0.091 F13.4-0.101 F31 + 0.013 o Bach factor in Equation (5) can have a value of 0 (unused) or 1 (used). There are many possibilities for cost overrun percen: tages between the maximum (133.8) and minimum (1.3) values. The main probability cases, including the (Min. and Max) val vues, were used to determine the cost overrun according to Equation (5). The proposed regression model is shown in Table 8. Figure 7 shows the CDF of cost overruns for the proposed sion model. The mean cost overrun of the model was ind to be 57.4228. The probability that the project will experi fence an average cost overrun of 57.422% as a percentage of the actual costs is les than 60%, Further, if the owner wants to 50% probability (Psa) as a contingency cost to sufficiently con. sider the potential cost overrun, then LL indicates that 47% of the cost contingency is required. 4.2, Model verification ‘As shown in Table 9, 12 projects were selected from the 75 pro} cts for practical testing of the model based on data and docu ments available from GARBLT, MHUC, final project statement, and real field supervision; ©All projects were open tender with unit price contract. ‘©The selection criteria were based on the projects’ region, and size to reflect road network projects in Egypt based on real data from these projects and actual practice # Data were collected from field investigation for these pro} fects from different sources such as contract agreement approved priced bill of quantities, approved drawings, and final statements Below are some detailed data on three of the twelve projects shown in Table 8; © Assiut-Sohag Tama axis (2009-2018) [P3] was estab- lished as a link between the eastern and western roads in Upper Egypt. The entire project consists of three phases: the first phase was completed in 2019, while the second and the third phases were completed in 2021 and 2022 respectively. The initial total cost of the project for the e phases was estimated at 1.55 billion LE, The studied case was the first phase, which started in May 2008, with a total length of 6.5 km and an initial cost of 300 million LE, and took 104 months to complete. © Regional Ring Road (2015-2018) [P30] is an oceanic road in Egypt that surrounds the Middle Ring Road, enclosing the ring road of Greater Cairo, The project started in 2015 and was launched on 9 September 2018, ‘The route passes through four governorates: El-Shargia, Qalyubia, Menoufla, and Giza, starting from Belbeis and ‘ending with the Cairo- Alexandria road intersection. ‘Zagazig-Sembelawin Road, phase 2 (2013-2019) [P71], this case study represents the construction of a dual road of a total length of 17.5km (Zagazig-Sinbillawain) as a traffic axis between the governorates of Shargia and Dakahlia, contributing to resolving the traffic jam between Zagazig and Sinbillawain, Figure 8 shows the Egyptian map, which includes the location, and governorate of the applied case studies As shown in Table 9, the model has been validated using sev- eral data from actual projects based on Equation (5). The valid ation indicates that the model reaches an overall accuracy of 83.64% compared to the provided real data,© tawma er aL ‘Cumulative Distribution Function Probabity Costovernun (Se) for Regression Model [= teprtoaistie GPF] Figure 7. The COF of cost ovens for the propose regression made Table 9. Comparison between cost over fr tal projects an model predcon Cost Overuns 0) Project Cont overrun factors Real projets BA model ay ‘Quantity anges 15) a733 Tae ral Scape of wark 18, quantity changes (15) 2126 36) (rin ‘Quarty hanges (3) 5631 100 (P23) Quantity changes (5), poor quality of work (4) sas 360 (P30) Scope of wark 18, quantiy changes (15) val er rss) ‘Quarty changes (8) an 1860 (ra) ‘Quantity ranges (5) a7 1860 (P44) ‘Quanity ranges (15) ais oO [P50] Scope of werk 18, quaniy changes (FI), nacurate cas estimate (F3) 7080 58 IPss) Scope of wrk (F18, Quantity ehanges (FS 2439 3700 (P63) Scope of wark F18, Quantity ehanges (FS nat 3670 Pn ‘Quanty ranges (FS), accurate east estimate (F13 3600 m7 Pverage He nn (G8) A model real projets new The average cost overrun % for the regression-based model ‘was 29.02%, covering more than 80% of the actual projects’ con: Lingencies, which is good enough to cover actual cost contin gency, and the model needs further improvement to get more accurate resulls, This may be achieved by getting more real data from several varieties of road network projects Inaccuracy in cost prediction by the developed model has been measured as the actual costs minus the predicted costs as a percentage of the predicted costs, An inaccuracy of ‘0’ means hat the project's expected cost is correct and equals the actual costs (Cantarelli and Fiyvbjerg, 2013). The accuracy and applic: ability of the proposed regression model for cost contingencies were verified using data collected from previous records for 75 RNCP implemented during 2000-2020. To verily the accuracy of the mean estimated contingency cost of 57.422%, as determined by the proposed regression model, the data were collected fom experts based on their previ fous real projects’ data, including the estimated and actual costs “The average contingency cost for the 75 projects under con: sideration was $9.722%, which is very close (96.15%) to the value produced by the developed model (57.4228). The proposed pre dicted contingency cost matched more than 96% of the mean catimated contingency of the actual projects, and the model pre diction matched more than 80% of real practice. Therefore, the developed regression-based model was reliable in predicting cost ‘contingency for RNCP in Egypt and could be validated for other countries. The practical and managerial implications of this research can be summarized as the developed model is expected to serve as a useful basis for road agencies to predict reliable cost contingency amounts, The model simplifies the calculations ‘of cost contingency for road network projects, thereby facilitating the incorporation of risk contingency for road projects. In add. ition, this methodology may be extended to other types of con- struction projects ‘The research contributes to predicting/calculating theoretical and practical - cost contingency percentages for RNCP, improv ing the decision-making process, and enabling construction agencies to manage their projects better. Conclusions ‘The cost performance of any construction project is a Key suc- cess criterion for the project owners, and it can be measured by ‘comparing the final cost to the project budget. Therefore, est ‘ating cost contingency is a crucial aspect of construction pro}- fects, Most road network projects experienced cost overruns between the contract award and the completion date, The scope ‘of work, quantity changes, improper verification of contract documents, and inaccurate cost estimates have been identified as the leading causes based on previous real case studies, Therefore,New Valley Figure 8. Egyptian governerate, Upper and North Egypt this research developed a simple regression model to determine the probability of incurred cost overruns based on the probabil distribution to estimate the appropriate construction contin gency costs, Considering the project price as a reference point, cost over: rns were calculated for 75 real RNCP from the Egyptian public sector under a unitprice contract, ranging from EGP 103 million Wo EGP 2.418 billion, Based on the results, the mean cost overrun, for the analyzed projects was 59.72%. No significant differences ‘were observed in the total cost overrun for all projects with dif: ferent price values, However, differences might occur according to the type and the size of the considered projects Non parametsic ‘goodness offi tests were used to select the ‘best-fit probability distribution. LL probability function (3P) sas used to provide the best overall distribution ft to calculate the probability of a cost overrun. It indicated that the fit of the empirical distributions was essential to achieve reliable and real istic probabilities of cost overrun. Results indicated that the expected contingency cost by the proposed regression model matched 96.15% of the mean est mated contingency for the selected sample projects and matched 83.64 of the average actual contingencies for the current practice projects, Therefore, the developed regression model is robust and reliable in predicting cost contingency values for RNCP in Egypt; however, more validations should be conducted for other countries. ‘These findings have significant implications for practitioners and researchers to improve future accuracy/caleulations for contingency amount for road construction projects, Using this approach to predict cost contingency percentages for RNCP will ‘enhance the decision-making process and enable owners to man- age their projects better, Based on the study analysis the outcomes corresponding to research questions have been summarized and can be listed as follow, Lhe tisk factors causing cost overruns have been shown in Table 2. 2. The study identified the common models that have been used by experts Simulations, Analytic Hierarchy Processes and Risk-Based Extimation, 3. The components of the proposed model to overrun and contingency percentages have been discussed in details in Section 4.1 4. The predictive accuracy of the proposed model has been dis- ‘cussed in Section 4.2. construction such as Monte Carlo edict cost All projects included in this study were unit-price contracts; therefore, no comparisons have been made among different con: tracting methods, Furthermore, this study ie limited to road net- ‘work projects in Egypt The data was collected from road network projects in Egypt. and these results may not reflect all the characteristics of such projects in other countries. "Thus, rmaore research can be conducted in other areas to increase the generalizability of these results ‘ature studies can consider14 © TAMAR ev AL different types of transportation projects, including railways, highways, airports, and seaports, for cost contingency predictions Disclosure statement No potenti conflict of interest was reported by the authors Funding ‘This research received no external fund Data availability statement ‘The authors confirm that all data supporting this study are available, References ‘Nb ELMnty A, HLKholy AM. Als AY. 2017 Maden shee nein ed ct ion perctaags af ghey projec og sy spend ECAML 243)s09-827, dot Iosl0w/scan03-2016-0084 AtShaaln AM 2020, Contingency secon and ranking for compote power sytem elablity elation. } King Sed Univ" Bag Se 320) Aub’. 202. Malle picewue regrenion for Wp generation modes ‘Aranda Bo J 10) 717-708. dt 101016} 309112073. Aimar I, Abdel Sonem Mi H'Dasb K.2022 Rk fore casing com vers in oud eth Ain Shame Hng | 13(5}101720. 101016, nds Ntahanads AM, Wang J. Zou PRO, Zhang, 2019, The cot evrmance aid nus of ete in ieee pro} ct in Asa Ch Hg Manag, 250}203-214 dat 10s¥ifee.2013 fete Asks J, 202, Regeson anal base. Regeion Ana Authony for Ronds snd Bdge and Land Transport (GARBLT), 200. ipso galego Ayub Thabeem Mf Uh F 2019. Contingency cele ding project, econ, the concer dncien mating denna Pijct May 5Ste}734-74, do 1011778756720 96a, scent D205. Banting prof can colingecy ~ beyond the 10% Syndrome, Austraan Inte of Poet Manageest”Caneene Vicor Auitlan nate of Pet Mangement acini D, Lave PED. 2014, Sali characteris co contingency in alr inastuctse peje} Conse Bag Manage. M403) da 1005 (Ge co 9s 702080. accra 200 cra inating poe ca cosrction contingency stitial ana Cobra: RIGS Ineraaona construction contre Reponding to dang, Hedaghy stadium. UX Beppe cuted sible 9001195779059 Castel CG Beg B 2013 Meg projec” cost pexormance and lok i probie and tone. Int Handbook on Megs Projet do 104337) Sratzaioonso8 an Cheng. C-L. Garp G. Shhh, CoetSclnt of determination for multiple ‘neuurement er mode | Mulvar Anal 2014 126137152 do It {OV ja 201401006 Ccucco , Wasteas Mf Juan G 221, The cede of determination F ‘ued is more ifomave than SMAPE, MAE, MAPE, MSE and RMSE sm epession analysis evalation. Per] Comput Set 7662, do 107717 xed Gy kor M, Wong IK. 200, rast of rk ator iening “eco oven im deiner f highway connsson poe, | Cont Eg Manage T(9)s28-597 dat. 10 oi/(ASCH)CO 1nd 786 0010, bykman A. Doan M, Baksh N.2018 Interpreting taonal ct on feng tehode inthe constuction indy Proveedingy of the {ttematona Confrence of Architectural Science Anociaton, 165-174 fot Toaesesigrn e772 "lasham ¥ Jone $019 Eaton and prediction of contin ot nde sing ber ewok in eri, and repesion Alemania Hog) 50} ‘oo Ste do Lo velshaq 209305002 12-50. dot 1, ELKholy AM 2015. Predicting cost overun in construction projects Tat Conse ng Manag. (4) 95-105, do 105923) em 2015040401 ELTouny AS, brahim AH, Amer ML 2014 Esumatng cos contingency for ighway consrucion projects using analytic Birtrchy procecte. Tot} [Compt Se Iseues. 166) 73-73. ELKholy AM Tabwaa AM, Hlayed MM, 2020, Prediction of simulated cost tuntingency for steal teinforceient in building projects” ANN. yerus fegression based modes. Int Constr Manag, 226) 1675-1689, do 10, Field A. 2020. Dieovering raises using TBM SPSS aatitie. Sth ed London: Sage lybjerg B Ansar A, Buduler A, Subl $,Canlarel C, Garbuio ML, Gentng ‘G, Holm MS. Lovalo D. Lunn D, eal 2018, Five things you should Know about cost overrun. Transport Rex, But A, Pohoy Pract 218174 190. doi 1010164201807 018 Gunhan S, Ardti D. 2007. Budgeting owner's construction contingency 1 Canste Eng Manag, 135(7)192-497. dot 101061 (ce)0733 5384 2007) 1337490), Hammad MW, Abbasi A, Ryan MJ. 2016, Allocation and management of| ‘cae contingency in projec] Manage Eng. 32(8) dot 101061 (aer)me, 943-5478,0000487, Herrera RE, Sinches O, Castaeds K, Porras #2020. Cost overren causative factors in road infrastructure projects frequency and importance ana Fate App Ses, 10(16)5508, dot 10.3390/app 10165508 tse UH, Maroul RG, Fabeem 1.2001. Anabye of rk core alcting the ‘main execution acts of roadways constuction projects. 7 King Saud Univ Bap Sel. 3563722383, dot 101016) hover 2021.05 00 Jckaon 2012 Contingeney for cot contol i project management: cae study. CEB. 34-12. dol 105130/A}CER wl 2906. ‘Karen JT, Lereim J. 2005. Management of proeet contingency and allow ‘ance Cost Bog. 470924. [hath BA, Poh YS, H-Shafie A. 2020. Delay factors management and rank ing for recontrution and rehabitalion projects based on the alive Importance index (RID. Sustainability 12(15)6171, doi: 103390) Lie 1, Zand N. 2010. "The accuracy of rk baeed cont extimation for water Iniratucture prokcts prchminary evidence from Ausalian projects Cones Manag Keon, 28(189-100. dos 10 1080/00448i90803451525. love PE tka LA. 2021, The ‘context of anspor project cost performance insights from contract award to Goal contracion costs, Res Transport eon, 90:101062, dot 10.1016) retre 2021 101062. Love PED, sing C-P, Carey 3, Kim JT. 2015. Eimating construction contin encp. accommodating the potential for coat overrne in rond contrvc ion projects. intastract syst 21(2). do 1006L/(aceis 948 555%, MarkizN, Irae A. 2022 tntegeaing an expert sytem with IMS cost et ‘ation, and lineat scheduling at conceptual design sage of Bridge pro) fete ot } Construct Manag. 22(3)912-928, dot 10 1080/25623599 2019, es1572 Minisuy of Housing, Uuites and Urban Communities. (MHUC). Ips anemic goeng ‘Morad AM 2003. 'An analysis of global constroction projects causes and ‘enpiations of cont overruns OSE Preprints. dt: 1031219704 iofsfyn6, Mure AH, Mall SM, Ngesa 0. 2021. On the loplogiic dntebution and 1 generalizations. a survey. HSP. 1063). do. 105539!spi0a3p95, ‘Oded. J. Welde M. 2021. Cost averrens of tansportaion infranteucture ‘projects, Int Encyclopedia Transport. 483-489. doi, 10 1016/6978-0-08, Inae7i-7.10089-2 ‘Odeck 2019. Vaviaton sn cost overruns of tansporaion projects an conometrc metarepression analysis of studies reported inthe leratur, Transportation 46(4:18%5~1366 dot: 101007/511116017 98365, Peleske CA, Dores V, Manian RA, Munteans R. 2015, Rk consideration {nd cost estimation in construction projets using Monte Carlo simula fon. Management 10(2) 163-175. Peter ED, Lave |, Smith I, Simpron M, Regan OO, 2015. Understanding the Tandscape of overruns in transport infasuucture projects Eavizon Pasa 1 Plann Des. 42(3)490-509, do. 10.1068/150102p. Rodhi NN. 2023. Comparison of probability impact matix and farzy logic methods for tsk amass of ral infrastructure projects in Indonel. Eg Technol J, 0805) dos 10 47191 es /ei5. 013, Segelod E2018. Project Cost) Overrun: causes, Consequences ‘Tovetinent Decisions, Cambridge: Cambridge Univeriy rest Shibani A, Tasan D. Saafan J, Sabboubeh Tf, Etaip M,Saidai M, Gherbal 1.2022, Financial sdk management sa the conssuction projects} King Snud Uni. Bing Se. do 101016) rues 2022 05.00, Smith NI, Mera T, Tobling P, Thompson S. 2014. Managing risk in com struction projects 3ed ed Chichester Wiley Blackwell sodSorendran , Tota Maharaj K. 2015 Log lois datibution to model water demand data, Procedia Eng 119:798-802. do 101016f} proeng 205 08 “Teeng ©, Zhuo TF C2009. Contingency etimasion uring a real options ‘pproach, Construct Manag Econ. 27(1) 1078-1087. dot 105080! ‘taasi9oos222411 World Bank Group, 2018. Undertending Cort Drivers of IMentiiation ‘Systems. Washington. D.C: The Wold Bank hep/elbrary workdbank otgdotvbook 10159531065. INTERNATIONAL JOURNAL OF CONSTRUCTION MANAGEMENT @) 15 ‘idiz 8, Bilbao I, Sproul AB. 2017, A review and analy of repression and ‘machin leaning models on commercial builing eect ond freeing, Renew Sut Energy Re 21104-1132 dot 10106) ue 2017 2023, ‘Yousry Alal A 2017. Mapping the exer of time, cost averrane and quality Shona in Egyptian publ highway projets, BEM. 36}120. det 10 {8 ebm 201703064 ‘Dorada J, Levitan A, Goan J. 2011. A comparison of regression and artifical tacllgence methods in a mass appraal copter. } Rea Estate Res. 333): 319-348. doe 10 1080/10835517 201112091311