You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.


Artificial Neural Networks in Construction Engineering and Management

Article · March 2017

DOI: 10.7492/IJAEC.2017.006


0 887

3 authors, including:

Baba Shehu Waziri Shehu Ahmadu Bustani

University of Maiduguri Spinal Engineering Services Limited, Abuja


Some of the authors of this publication are also working on these related projects:

Risk factors influencing cost overrun in construction projects View project

Green Construction in Nigeria View project

All content following this page was uploaded by Baba Shehu Waziri on 22 June 2017.

The user has requested enhancement of the downloaded file.

International Journal of Architecture, Engineering and Construction
Vol 6, No 1, March 2017,50-60

Artificial Neural Networks

in Construction Engineering and Management
Baba Shehu Waziri1,∗ , Kabir Bala2 and Shehu Ahmadu Bustani3

Department of Civil and Water Resources Engineering, University of Maiduguri, Nigeria
Department of Building, Ahmadu Bello University Zaria, Nigeria
Department of Building Technology, Abubakar Tafawa Balewa University, Bauchi, Nigeria

Abstract: Artificial Neural Networks has gained considerable application in construction engineering and management in recent
time. Over 100 resources published in refereed journals and conference proceedings were screened and reviewed with the view to
exploring the trend and new directions of the applications of different ANN algorithms. The study revealed successful applications
of ANNs in cost prediction, optimization and scheduling, risk assessment, claims and dispute resolution outcomes and decision
making. It was observed that ANN have been applied to problems that are difficult to solve with traditional mathematical and
statistical methods. The integration of ANN with other soft computing methods like Genetic Algorithm, Fuzzy Logic, Ant Colony
Optimization, Artificial Bee Colony and Particle Swarm Optimization were also explored which generally indicated better results
when compared with conventional ANNs. The study provides comprehensive repute of ANN in construction engineering and
management for application in different areas for improved accuracy and reliable predictions.

Keywords: Artificial Neural Network, construction management, construction engineering, construction cost estimation, con-
struction scheduling


1INTRODUCTION be traced back to 1989. Subsequently, they have successful-

ly been applied to solve numerous problems in engineering and
Artificial Neural Networks (ANNs) are computational mecha- management. In the field of construction management they have
nisms that have the ability to acquire, represent and compute been applied for tender price prediction (Li and Love 1999), con-
function from one multivariate space of information to another struction cost estimation (Williams 1994;Emsley et al. 2002;
given a set of data representing that function. ANNs are func- Wilmot and Mei 2005;Sodikov 2005;Pewdum et al. 2009;Alex
tional abstraction of the biological neural structure of the central et al. 2009;Waziri 2010;Bala and Waziri 2012;Bala et al.
nervous system that are more effective than traditional methods 2014), project cash flow (Boussabaine and Kaka 1998), produc-
for solving complex qualitative or quantitative problems where tivity forecast (Chao and Skibniewski 1994;Portas and AbouR-
the parameters for conventional statistical and mathematical izk 1997;Boussabaine 1995;Savin et al. 1998;Al-Zwainy et al.
methods are highly interdependent and data is intrinsically noisy 2012), dispute resolution (Yitmen and Soujeri 2010;Fatima et al.
or incomplete or prone to error (Rumelhart 1986;Adeli 2001; 2014), earth moving operation (Shi 1999), contractors prequalifi-
Aleksander and Morton 1993;Rudomin et al. 1993;Arbib 1995; cation (Lam et al. 2001), contract performance (Zin et al. 2006;
Geon 2005;Sivanandam and Deepa 2006;Bala et al. 2014). Waziri 2012), mark up estimation (Li and Love 1999) risk quan-
ANNs have been established to be powerful pattern recognizers tification (McKim 1993;Maria-Sanchez 2004;Wang and Elhag
and classifiers which operate as a black box to learn significant 2007;Xiang and Luo 2012;Liu and Guo 2014); time contin-
structures in data (Adeli 2001;Jain and Pathak 2014). They gency (Yahia et al. 2011). Inspite of the numerous advantages
are composed of a large number of highly interconnected pro- of ANN (such as adaptive learning, Self-organisation, Real time
cessing elements called neurons working in unison to solve spe- operation and Fault tolerance) over traditional statistical tool,
cific problems. ANNS are fundamentally characterised by their yet it offers little explanation on the relationships between the
architecture (connection between neurons); training or learn- parameters used for modelling which makes it difficult to explain
ing (determining the weights on the connections) and activation what is learnt from the network (Paliwal and Kumar 2011). It
function. is therefore envisaged that further research into the framework
According to earlier application of ANN to engineering can and internal process within the neural network will offer better

*Corresponding author.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

explanatory insight into the influence of independent variables based on relevance and significance toward understanding and
in the modelling process. documenting the trend and new direction of ANN application.
A vast application of ANN in the fields of construction Engi-
neering and Management for solving crucial construction deci-
sions are based on the simple back propagation algorithm. The
Back Propagation (BP) training algorithm is the most popu-
lar typology and learning method. Several other neural netw-
orks other than the BP such as the regularization neural network ANN have been successfully applied to predict tender price
had been developed to deal with noise and over-fitting prob- (McKim 1993;Li and Love 1999), construction cost pre-
lems in data. The typical architecture of the feed forward Neu- diction (Williams 1994;Emsley et al. 2002;Wilmot and Mei
ral Network illustrated in Fig.1consists of an input layer, hid- 2005;Sodikov 2005;Pewdum et al. 2009;Alex et al. 2009;
den layers and output layer. The neurons in the input layer are Waziri 2010;Bala and Waziri 2012;Bala et al. 2014), project
connected to those in the hidden layers by the synaptic weights. cash flow (Boussabaine and Kaka 1998), Labour productivity
The common transfer functions used are the summation function (Chao and Skibniewski 1994;Portas and AbouRizk 1997;Savin
and the sigmoid squashing function. et al. 1998), earth moving operation (Shi 1999), contractors pre-
qualification (Lam et al. 2001), contract performance (Zin et al.
2006;Waziri 2012), mark up estimation (Li and Love 1999) risk
quantification (McKim 1993).

Cost Estimation

Construction cost estimation is a crucial activity for proper func-

tioning of any construction firm (ElSawy et al. 2011). The
application of ANN in cost estimation has been the subject of
many studies (Pearce 1997;Bhokha and Ogunlana 1999;Son-
mez 2004;Sodikov 2005;Kim et al. 2005;Cheng et al. 2009a;
Cheng et al. 2009b;Arafa and Alqedra 2011;Waziri and Bala
2011;Bala et al. 2014).
Williams (1994) used the BP algorithm for predicting changes
in construction cost indices for one and six months ahead and
concluded that the movement of the cost index is a complex
problem that is difficult to be predicted accurately using the BP
model.Hegazy et al. (1994) used ANN application for optimum
mark-up estimation and discussed its potential applications in
construction Engineering and Management.Hegazy and Ayed
(1998) developed a parametric cost estimating model for high-
Figure 1. Feed forward neural network
way projects based on the ANN approach. In the study, two
The first mathematical representation of neuron (processing alternative techniques namely Simplex optimization and Genet-
elements of the network) was attempted in 1943 by the Neu- ic Algorithm (GA) were introduced to train the network weights.
ro physiologist Warren McCulloch and the Logician Walter Pits Adeli presented a regularization neural network model for esti-
(Galkin 2002). The representation is shown in Fig.2. The mating the construction cost of reinforced concrete pavement
McCulloch and Pitts (MCP) neuron is binary activated. projects. In the study they observed that highway construction
costs are noisy due to the multiplicity of interplaying factors fluc-
tuations, weather conditions and human judgement resulting in
over-fitting. The RNN model proved to be pragmatic for reli-
able and consistent cost estimation of highway projects.Geiger
et al. (1998) developed a model to estimate the cost of sheet
metals from direct material cost and cost of supplied parts by
the use of a ANN. The results showed that within the investi-
gated range, an accuracy of 5% to 15% was achieved.Elhag
and Boussabaine (1998) presented two ANN models to predict
the lowest tender price of primary and secondary school build-
ings using data of 30 completed projects for training the net-
Figure 2. Artificial neuron works. The results revealed that the two models effectively
learned during training stage and gained good generalization
The paper reviewed literature form well over 100 resources capabilities in training session resulting in prediction accuracy
published in refereed journals and conference proceedings with of 79.3% and 82.2%.Al-Tabtabai et al. (1998) developed an
the view to exploring the trend and directions for the appli- ANN model for the prediction of percentage increase in the cost
cations of different ANN algorithms in construction engineer- of typical highway project from a reference estimate. The model
ing and management. The resources were screened and selected achieved a MAPE of 8.1%.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

Bhokha and Ogunlana (1999) in their study for developing gression model were established for the study. One of the NN
an ANN model for predicting the construction costs of building models achieved an accuracy level of 12% which was considered
projects in Thailand at the pre-design stage used historical da- satisfactory for conceptual cost estimating.Kim et al. (2004)
ta of 136 completed properties. The cost variables used in the examined the performances of multiple regression analysis, neu-
study as inputs are: structural system, building function, exteri- ral network and cased based reasoning in prediction of building
or finishing, building height, decorating class and site accessibil- project cost. The results revealed that the neural network mod-
ity. The validation results indicated a satisfactory performance el performed the best prediction in terms of accuracy but cased
where on the overall 42.7% of the sample were underestimated based reasoning indicated better performance in the long run.
while 57.3% were overestimated.Fang and Froese (1999) used Günaydın and Doğan (2004) proposed a neural network model
neural network approach to establish relationship between the for cost estimation of structural system of buildings where the
qualities, cost of concrete and formwork for the structural ele- model achieved an accuracy level of 93%. They concluded that
ment of tall buildings using high performance concrete (HPC). neural networks are capable of reducing the uncertainties of es-
Hybrid and hierarchical neural networks were used to predict the timate for a structural system of building.Kim et al. (2004)
quantities/cost of HPC wall frame structures in tall commercial compared the prediction ability of ANN, CBR and regression
buildings. The results of the comparison of the two strategies analysis based on a historical cost data of 530 Korea residential
revealed that the hybrid model is less accurate but easy to be construction projects. The results demonstrates the potentials of
trained while the hierarchical models are more accurate but more modelling with ANN obtain more accurate results as opposed to
complicated in implementation. Both of the strategies were ob- CBR and regression analysis.Wilmot and Mei (2005) employed
served to provide promising results.Shtub and Versano (1999) ANN models which related overall highway construction co-
proposed a system to estimate the cost of steel pipe bending us- sts described in terms of highway construction cost index to the
ing ANN and regression analysis. The results of the evaluation cost of construction of materials, labour and equipment. The
of the models revealed that the neural network model outper- study indicated that the model was able to replicate past high-
formed the linear regression model by prediction performance way construction cost trends in Louisiana with reasonable ac-
but the regression model has the best fit of the data.Assaf curacy.Sodikov (2005) examined cost estimation for highway
et al. (2001) used ANN to investigate the overhead cost prac- projects by employing ANN and observed that neural network
tices of construction firms in Saudi Arabia. The proposed model is an appropriate tool for solving complex problems and can
would be used by construction firms to decide an optimum lev- also cope with imprecise data. The results demonstrate good
el of overhead costs that enables them to win and effectively nonlinear approach ability and higher prediction accuracy of
administer capital projects. back propagation neural network. Sayed andIranmanesh and
Emsley et al. (2002) compared the prediction performances of Zarezadeh (2008) presented the application of ANN to forecast
regression analysis and ANN based on a dataset of 288 proper- actual cost of construction projects based on the Earned Value
ties. The study considered 41 independent variables including Management system (EVMS) to reduce the risk of project cost
site related variables and design related variables. The results overrun. The model was evaluated by the MAPE criterion which
showed that in the best case, the model indicated a Mean Ab- showed satisfactory performance.Bouabaz and Hamami (2008)
solute Percent Error (MAPE) of 17%. This error term is too proposed a model for estimation of repair and maintenance of
large to enable the practical application of the model. In ad- bridges in developing countries based on the ANN technique for
dition, the necessary input variables were extensive, making it better accuracy. Cost and design data for two categories of re-
difficult to apply in early design stage.Setyawati et al. (2002) pair bridges were used for training the network model which
used neural network to develop cost estimating model for insti- achieved an accuracy level of 96
tutional buildings and obtained an accuracy of 16%.Pathak and Sonmez and Ontepeli (2009) employed regression analysis and
Agarwal (2003) proposed a programme based on ANN for de- ANN for developing parametric models for estimating construc-
sign, estimation and costing of intz and circular overhead water tion cost of urban railway system. Two neural networks incor-
tanks in Bhopal region of India. The input parameters for the porating different independent variables were considered as an
programme are height to diameter ratio, angle of conical wall, alternative to regression model for the identification of the non-
number of columns. The trained network predicted the cost of linear relations. The performance evaluation of the models re-
new tanks with a 3.16% error margin which indicated a satis- vealed that one of the NN models provided the best results in
factory performance. Such quick and reliable cost prediction of terms of accuracy.Wang et al. (2010) employed BP neural
water tanks will be helpful in the selection of tanks for design network for estimating highway projects costs. The model was
and construction purposes.Günaydın and Doğan (2004) devel- trained by a dataset obtained from some successful highway en-
oped an ANN model for estimating cost of structural systems of gineering projects to provide quick cost estimating. The results
reinforced concrete skeleton buildings in Turkey. Cost and de- indicated the practicability and reliability of the model posing
sign data of thirty (30) projects with eight parameters were used promising prospects of BP-NN for cost estimating of highway
in training and testing the ANN methodology. An average cost engineering construction.
estimation accuracy of 93% was achieved. The model is useful Arafa and Alqedra (2011) employed ANN to develop an effi-
for design professionals to make appropriate decisions at early cient model to estimate the cost of building construction projects
project phase. at the early stage. The study used datasets of 71 building
Sonmez (2004) compared regression and neural network mod- projects in Gaza strip. Significant parameters obtainable at the
els for conceptual cost estimation using construction year, lo- pre-design stage were used as input variables for model develop-
cation index, proportion car parking area and area for comm- ment. The results of the study indicated that ground floor area,
ons as the independent variables. Two NN models and one re- number of storeys, types of foundation and number of elevators

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

in the building are the most effective parameters influencing ear- sion model respectively.Yadav et al. (2016) developed a cost
ly stage estimates of building cost.ElSawy et al. (2011) present- estimating technique based on the principles of ANN to forecast
ed a BP neural network model for the prediction of site overhead structural cost of residential buildings. Twenty three years data
cost in Egypt. The study used data of 52 real life projects exe- were collected from schedule of rates records for training and
cuted between 2002 and 2009 for training whereas five (5) new testing of networks. The parameters collected included, cost of
projects data were used for the validation. The results indicated cement, sand, steel, aggregates, mason, skilled and non-skilled
a Root Mean Square Error (RMS) value of 0.2764 and an accura- worker. The parameters were simulated using NEURO XL ver-
cy level of 80%. The model was observed to predict wrongly the sion 2.1. The neural model predicted total structural cost of
percentage of site overhead costs for only one project (20%) of building projects with correlation coefficient R of 0.9960 and R 2
the testing sample.Ahiaga-Dagbui and Smith (2012) employed value of 0.995.
ANN to model the final target cost of water projects in Scotland
based on the data of 98 water related projects executed between
2007 and 2011. Different models were developed for normal-
Construction Scheduling
ize target cost and log of target cost variable transformations
and weight decay regularization were then explored to improve
the final models performance. The investigation revealed that Adeli and Karim (1997) applied a general mathematical formula-
ANN was able to capture the interactions between the predictor tion for the problem of highway construction scheduling. A neu-
variables and final cost. ral dynamic model was employed to solve the non-linear prob-
Vahdani et al. (2012) presented a computationally effi- lem with the goal of minimizing the direct construction dura-
cient model called the support vector machine (SVM) to im- tion. The model provides the capabilities of both Critical Path
prove the conceptual cost estimating accuracy during the early Method (CPM) and linear scheduling approach yielding opti-
phase of project lifecycle. The model was trained by a cross mum schedule with minimum cost. This methodology is consid-
validation technique and its performance results were compared ered satisfactory for solving cost-duration trade-off problem of
with those of non-linear regression and BP-NN which revealed highway construction. The study also provides foundation for
that the SVM had the best results. The work ofAlqahtani and development of a new generalization of more flexible and accu-
Whyte (2013) also employed ANN technique to develop a new rate construction scheduling systems.Graham et al. (2006) used
framework for life Cycle Cost Analysis (LCCA) of construction ANN for predicting duration of Ready Mixed Concrete (RMC)
projects. The model computes whole life cycle costs of construc- which is assumed to be seriously related to construction opera-
tion projects using the cost of significant items (CSI) to identify tions. The study used data obtained from four different projects
main cost items. MATLAB and Excel solver were used to de- consisting of the variables; months of operation, type of opera-
velop the models using a dataset of 20 building projects. The tions, total operation volume, average inter-arrival time, num-
results revealed accuracy levels of 1% and 2% for the MATLAB ber of loads, truck volume and number of rejected loads.Yahia
and Excel solver respectively. et al. (2011) employed ANN to develop a model for a more re-
Lyne and Maximinio (2014) developed an ANN model for the liable prediction of the amount of time contingency that should
prediction of total structural cost of building projects in Philip- be added to the scheduled project completion time.Petruseva
pines using historical data of 30 completed projects. The da- et al. (2013) presented a supervised learning algorithm called
ta was randomly divided into 60% for training, 20% for vali- the support vector machine (SVM) for predicting construction
dating the performance while the remaining 20% as complete- duration. Contracted and real price data of 75 building construc-
ly independent test for network generalization. Six input pa- tion projects initiated and completed between 1999 and 2011 in
rameters namely; number of basements, floor area, number of the Federation of Bosnia Herzegovina were obtained through
storeys, volume of concrete, area of formwork and weight of field survey for analysis. The study used regression analysis and
reinforcement steel were used. The results showed that ANN SVM network to achieve improvement in the accuracy of project
model reasonably predicted the total structural cost of building duration prediction. The results indicate that predicting with
projects with favourable training and testing phase outcomes. SVM was significantly more accurate.
Kim et al. (2004) investigated different parametric cost esti-
mating techniques for construction projects and discovered that Maghrebi et al. (2014) used ANN to predict the duration of
neural networks generated better results than CBR and multi- a concrete operation by focusing on supply chain parameters of
ple regression analysis.El-Sawah and Moselhi (2014) employed RMC. The model was tested with a real life dataset of a RMC
Back Propagation Neural Network (BP-NN), Probabilistic Neu- in Sydney metropolitan area which has 17 depots and around
ral Network (PNN) and Generalized Regression Neural Network 200 trucks. The results obtained compared favourably with re-
(GRNN) and regression analysis models for order of magnitude sults from other studies that only considered the construction
cost estimating of low-rise structural steel buildings and their parameters for predicting productivity of concrete.Golizadeh
respective cost. The results of the investigation revealed that et al. (2016) proposed a tool for estimating duration of major
the MAPE of the neural network models ranges from 16.83% activities relating to the structural elements of concrete frame
to 19.3%, whereas for the regression model it was found to be buildings. Four AN models were develop to compute the du-
23.72%. The linear regression model was more sensitive to the ration of installing column reinforcement, beam reinforcement,
change in number of training data while the PNN was the most column concreting and beam concreting activities. Then a web
stable network among all the three models with maximum dif- based programme was developed as an automated tool for esti-
ference in MAPE of 2.46%. The maximum difference in MAPE mating duration based on the ANN for more accurate activity
was 19.47%, 17.91% and 61.45% for BPNN GRNN and regres- duration prediction.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

Decision Making Elhassan et al. (2012) discussed the possibilities of address-

ing the difficulties in decision making in construction manage-
Kamarthi et al. (1992) employed two layer BP neural network ment by the application of optimization techniques. The study
for the selection of formwork systems. The study indicated a identified optimizing tools being a basis for making an optimal
satisfactory performance.Murtaza and Fisher (1994) presented decision in respect of construction project management. They
an ANN model which enables decision making on using modu- concluded that there are quite some research gaps in the use
larization or conventional method for building an industrial pro- of methodologies for optimum decision making and pointed out
cess plant based on five categories of decision attributes namely the potentials of artificial (AI) such as ANN for future stud-
plant location, environmental and organizational factors, labour ies.AbouRizk et al. (2001) used a two-stage neural networks
related factors, plant characteristics and project risks. The NN analysis for the estimation of labour productivity rates for in-
model was trained using cases collected from several engineering dustrial construction activities. The method predicted with an
consulting and client firms.Boussabaine (1995) developed a neu- accuracy of 15%.Sawhney and Mund (2001) applied ANN based
ral network system for forecasting productivity and construction on the Bayesian classifiers method for the selection of type and
cost. The model proves the feasibility of an integral knowledge model of crane. The model exhibited a satisfactory performance
based system for construction planning and productivity. in the selection of crane based on their type and model.Lou
Masri et al. (1996) presented an ANN approach for detecting et al. (2001) employed ANN to predict the short term future
changes in the characteristics of structure of unknown systems. conditions of pavement cracks based on past conditions records.
The neural network was trained for identification using vibra- The model demonstrated the potential of applying ANN for such
tion measurements from a healthy system. The trained network predictions with satisfactory results.Morcous (2002) made com-
was fed with comparable vibration measurements from the same parison between CBR and NN in modelling bridge deterioration
structure under different conditions of response in order to mon- based on bridge deck data obtained from Ministry of Transporta-
itor the health of the structure. The study revealed that the tion of Quebec to compare the advantages of two methods to
proposed methodology is capable of detecting relatively small guide transportation agencies in selecting the most appropriate
changes in the structure parameters even if the vibration mea- approach.
surements are noisy.Pearce (1997) used ANN for cost based Chew and Tan (2003) presented a maintainability grading sys-
risk prediction and identification of project cost drivers. The tem using ANN which aids in enhancing decision-making of wet
study investigated the effectiveness of ANN to predict risk re- areas design. The model was derived from comprehensive con-
lated to final project cost and to identify potentially significant dition surveys of 450 tall buildings and interviews with relevant
cost drivers relating to construction projects. Neural network building professionals. In the study, 16 significant risk factors
models were developed over a set of permutations of input vari- were identified and tested according to their sensitivity in affect-
ables and used to generate a maximum cost versus probability ing maintainability scoring of wet areas. The system provides
curves which can be used to evaluate risks of cost growth between for complete evaluation of various alternative designs, construc-
conceptual design and project completion. The cost variables tion, materials and maintenance practices so as to achieve best
identified in the study includes floor to floor height, external possible solutions of technical attributes that lead to minimum
wall area, exterior window area, number of floors among others. lifecycle maintenance cost.Al-Sobiei et al. (2005) used ANN
Results of the research further indicated that ANN can serve and GA to predict the risk of contractor default in construction
as a robust tool for approximated multivariate analysis.Chua projects undertaken for the Saudi Arabia Armed Forces. The
et al. (1997) proposed an ANN model for the identification of study is useful in making a decision to engage the services of
key measurement factors that affect budget performance in a a contractor. The outcome of the research is of importance to
project. The technique has the capabilities of modelling even clients and other sponsors of construction projects because it
if the functional interrelationships between input factors and proposes an approach that can allow the use of a rational and
output performance could not clearly be defined. In the study, effective policy.Apanaviciene and Juodis (2005) applied NN to
eight (8) key variables were identified covering aspects related develop a model for predicting construction project management
to project manager, project team, and planning control efforts effectiveness from the perspectives of construction management
viz; scope, completion of design, number of organizational lev- organizations. Performance data from construction management
els between project manager and craftsmen, experience on simi- companies consisting of twelve key factors in Lithuania and the
lar projects, constructability programme, project team turnover USA were used for model development. The study recommends
rate, frequency of budget updates, frequency of control meet- the Construction Management Performance Evaluation Model
ings during construction and control system update. The model (CMPEM) as a decisionĺCsupport tool for competitive bidding
could be used to predict different management strategies to ef- and for evaluating management risk of construction projects.
fectively deploy resources to strengthen project management. Zin et al. (2006) presented an ANN model for predicting the
Yeh (1998) demonstrated the possibilities of adapting ANN to time performance of traditional contract projects. Several neu-
predict the compressive strength of High Performance (HPC) ral network models were developed and tested using nine sample
concrete. A set of trial batches produced in the laboratory were projects data. The best model for the prediction is a MLP, BP
used for the training and testing of the models. The results re- network with eight input nodes, five hidden nodes and three out-
vealed that ANN model is more accurate than a model based put nodes with a very low error of prediction.Golpayegani and
on regression analysis. The study also indicated that it is con- Emamizadeh (2007) presented a framework for planning work
venient and easy to use ANN models for numerical experiments breakdown structure of construction projects based on ANN.
to review the effects of the properties of each variable on the The approach uses the project control work breakdown structure
concrete mix. (PWBS), functional work breakdown structure (FWBS) and rel-

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

ative work breakdown structure (RWBS) to form the output of survey conducted to contractors in North Cyprus which formed
the model and its modules. The framework was tested on a the basis for model development. The model manages change
sample domain and the results showed that the planned work orders through all phases of project such that construction op-
breakdown structure and activities have satisfied the expecta- erations can continue with the least amount of interruption that
tions with different levels of validity.Khalafallah (2008) pre- usually results from disputes between different parties involved
sented an ANN based model for predicting housing market per- in a project. The data for the study was obtained from 35 cas-
formance to support real estate investors and home developers. es collected from 22 building contractors comprising of 11 input
The study used historical market performance dataset for train- variables and their corresponding influence on performance. The
ing the NN in order to predict unforeseen future performances. proposed model has been observed to be an efficient approach
The validation results revealed prediction in the range of -2% and to find the probability of dispute in respect of the identified pa-
+2%.Jamil et al. (2009) demonstrated the possibility of adapt- rameters.
ing ANN in the development of simulator and intelligent sys- Chou (2012) presented a model for predicting dispute handling
tem for the prediction of compressive strength and workability methods in Public-Private-Partnership (PPP) projects. The
of high performance concrete (HPC). The model demonstrated study used machine learning (ANN, SVM and Tree Augment-
satisfactory ability in learning the given input/output patterns ed Name (TAN), Bayesian), classification and regression based
indicating the appropriateness of the application of ANN in the techniques (classification and regression tree (CART), Quick Un-
field of HPC mix design. biased and Efficient Tree (QUEST), Exhaustive Chi-square Au-
Aibinu (2011) proposed a learning algorithm based on the tomatic Interactive Detection (Exhaustive CHAID) and C5.0
characteristics of completed projects for the quantitative and and combination of these methods for possible better perfor-
objective estimation of the inaccuracies in pretender cost esti- mance for a set of PPP data. The results showed that the com-
mates of new projects. A three layer feed forward ANN model bination of the techniques of QUEST + CHAID + C 5.0 demon-
was developed and trained to generalize nine characteristics of strated best classification accuracy at 84.65% in predicting dis-
100 completed projects. Nine input variables namely, project pute resolution outcomes (mediation, negotiation, arbitration,
size, procurement route, project type, project location, princi- litigation, adjudication, appeals or no dispute occur). The
pal structural material, sector, estimating method and estimated CART model revealed the best classification with accuracy of
sums were used. The model had a correlation coefficient (R) of 69.05%. The study demonstrates effective classification appli-
73%, Mean Absolute Error (MAE) of 3% and Mean Square Error cation for early project dispute resolution related to public in-
(MSE) of 0.2.Al-Zwainy et al. (2012) develop BP NN model frastructure projects.Fatima et al. (2014) identified significant
for construction productivity estimation of finishing works for qualitative parameters and used to develop an ANN model for
floors with marble. The study considered residential, commer- minimizing construction disputes and reduce the cost of con-
cial and educational projects data from different parts of Iraq. struction by optimizing the identified parameters. The method-
Ten (10) key factors including age, experience, number of the ology integrates the concept of ANN with the current estimating
assist labour, height of floor, size of the marble tiles, security system and optimizes the frequency of occurrence of dispute pa-
conditions, health status for the work team, weather condit- rameters which in turn reduces the cost of the project.
ions, site conditions and availability of construction mater-
ials were used as input variables. The results showed that Risk Assessment
ANN has the ability to predict the productivity for finishing
work with coefficient of correlation of 87.55% and prediction ac- Odeyinka et al. (2002) attempted to model the variation be-
curacy of 90.9%.Aswed (2016) employed ANN for the prediction tween predicted and actual cost flow due to inherent risk in
of labour productivity based on thirty influencing factors in Iraq. construction. They employed BP neural network to develop a
The factors used as input variable include, age, experience, gang cost flow risk assessment model. The model was tested on 20
health, gang number, wages, weather, material availability, site new projects with satisfactory prediction of variation between
conditions wall length, wall height, mortar type, wall thickness forecast and actual cost flow at 30%, 70% and 100% stages.
and site security. The model predicted actual labour produc- Maria-Sanchez (2004) employed neural network approach to as-
tivity with a reasonable degree of accuracy with coefficient of sess the impact of environmental risk in construction projects
correlation of R = 86.28%. The study concludes that the model in Puebla, Mexico. The network was trained and tested with
can be employed to predict labour productivity of any build- data obtained from private contractors that are constantly in-
ing type using the influencing factors.Sharmik et al. (2016) volved with projects facing environmental risks. The methodolo-
presented a cost and time effective feed forward BP-NN with gy demonstrates the potential of ANN in evaluating environmen-
supervised learning algorithm for estimating soil characteristics. tal risks and providing valuable outcomes for project managers
The model revealed a satisfactory results when compared with working with government agencies. The system also offers a con-
actual values of soil characteristics. siderable advantage in predicting the possible value of the total
environmental risks.Wang and Elhag (2007) compared the mod-
Dispute Resolution and Litigation elling mechanisms of Neural Network (NN), Multiple Regression
Analysis (MRA) and Evidential Reasoning (ER) and evaluated
Yitmen and Soujeri (2010) presented an ANN model for the esti- their performances using a set of bridge risk data. The study
mation of the influence of change orders on project performance revealed that ANN had better performance over ER, and MRA
for avoidance or resolution of disputes before litigation occurs. for the case study considered.
Significant factors that describe the adverse effects of change Xiang and Luo (2012) proposed a principal project parties’
orders on project performance were identified through a field behavioural risk evaluation model based on BPĺCNN. The BP

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

was employed to avoid subjectivity factors in the risk evaluation sults than simple BP neural networks. The model provides an
process. A likert scale of 1-5 was used to assess the risk factors option of whether or not to take a case to litigation.Cheng et al.
identified through field survey. The network simulation results (2009a) presented a method combining three different soft com-
show that the model is satisfactory and practical.Polat (2012) puting techniques namely, genetic algorithms, fuzzy logic theory
also proposed a contingency estimation model based on ANN to and neural networks under a mechanism referred to as Evolu-
enable managers assess the risk level of their projects in a more tionary Fuzzy Hybrid Neural Network Model (EFHNN). The
objectives and systematic manner thereby allowing them to es- proposed mechanism was developed for design phase cost esti-
timate cost contingency amount more reliably and accurately. mation of projects in Taiwan. The approach incorporates neural
Training and testing data were obtained from the records of 195 networks and high order neural networks (HNN) which operates
completed international projects undertaken by 85 large-scale with the alternative of linear and nonlinear neuron layer con-
contractors in Turkey. Statistical analysis of the results indicat- nectors. The approach also incorporates fuzzy logic for handling
ed that the model is valid and captures significant components of uncertainties. The approach therefore evolves fuzzy hybrid neu-
the underlying complex nonlinear relationship between the risk ral network (FHNN). For the optimization of the FHNN, GA is
factors and contingency amount included in the bid price. used which resulted in EFHNN. The model achieved an overall
Lhee et al. (2014) presented a two-step neural network based estimate error of 10.36% due to the use of GA, the method has
method for estimating optimal contingency for transportation a high computing time, this being a disadvantage.Cheng et al.
construction projects. The model provides the owner with opti- (2009b) presented a web based hybrid model incorporating ge-
mum solution with the view to improving budgeting decisions, netic algorithms, fuzzy logic theory and neural networks under
reducing the risk of either underutilizing or over committing of a mechanism called Evolutionary Fuzzy Neural Inference Model
funds.Liu and Guo (2014) constructed a risk evaluation model (EFNIM). However, EFNIM also runs long time due to the use
of project construction quality on the basis of neural networks of GA.
and rough sets. A dataset of residential building projects in the The study ofShi and Li (2010) integrates the use of fuzzy log-
Guangzhou development zone were used to test the model ac- ic, PSO and ANN in quality assessment of construction projects.
curacy employing research tools of Rosetta based on rough sets In the study, fuzzy logic was used to define the elements of an
and MATLAB 7.0. Empirical results showed that the model has assessment matrix and a quality assessment model for construc-
great practical significance. tion was set up. PSO was adapted to train the perception in the
assessment and predicting the quality of construction projects in
Evolutionary Neural Network in Construction china. Comparing BP-NN and ANN based on GA, the simulat-
Engineering and Management ed results of quality assessment of construction projects shows
that training the network with PSO gave more accurate results
Yeh (1998) employed Simulated Annealing Neural Network in terms of Sums of Squares Error (SSE) and faster in terms of
(SA-NN) to optimize construction site layout. SA is a probabilis- number of iterations and simulation time than the BP-NN and
tic hill-climbing search algorithm which can find a global min- GA-NN.Feng and Li (2013) presented an optimization method
imum of the performance function by combining gradient des- for cost estimation by integrating GA and BP technique. Eigh-
cent with a random process. This algorithm combined with ANN teen project cases and two testing samples were used to observe
demonstrates a satisfactory performance.Kim et al. (2004) em- generalize ability of the model. Comparing with conventional
ployed BP neural network model and Genetic Algorithm (GA) BP models, the study revealed that the GA-BP model can get
for cost estimation in a technique referred to as BP-GA. The lower forecast error and iterations but runs long time. They
GA was introduced in the study to improve the accuracy of the concluded that the model is appropriate for construction cost
BPN. Cost data of 530 residential buildings constructed in Korea estimation.
between 1997 and 2000 were used for training and performance Hong et al. (2014) put forward a construction engineering cost
evaluation. The hybrid BP-GA model was found to produce evaluation model and application based on RS-IPSO BP neural
effective and more reliable results compared to the BP model network called “the model of construction engineering cost e-
based on trial and error (Kim et al. 2004). valuation of optimized particle swarm and BP neural network”.
Kim et al. (2005) presented a hybrid model comprising of PSO was adopted to optimize the initial weights and threshold
ANN and Genetic algorithm (GA) for the estimation of pre- of ANN. The main aim of the hybrid method is to improve the
liminary costs of residential buildings. Residential construction rate of convergence of ANN and the ability to search for global
data initiated and completed between 1992 and 2000 in South optimum. This method is considered to have a high practical
Korea were used for training. Comparison between actual and value and it can be applied to make scientific evaluation of con-
predicted results showed that the mean, standard deviation and struction costs.Kayarvizhy et al. (2014) compared the improve-
the coefficient of determination (R2) of the ratio between ac- ment in the prediction accuracy of ANN when it is trained using
tual and predicted costs are 0.960, 0.420 and 97% respectively. swarm intelligence algorithms. Several models were formulat-
The results confirmed the ability of GA in overcoming the prob- ed for evaluating the various ANN-swarm intelligence combina-
lem of lack of adequate rules for determining the parameters of tions. The swarm intelligence algorithms considered in the s-
ANN.Chau (2007) applied Particle Swarm Optimization (PSO) tudy are Particle Swarm Optimization (PSO), Ant Colony Op-
based ANN in the analysis of outcomes of construction claims in timization (ACO), Artificial Bee Colony (ABC) and Firefly. The
Hong Kong considering cultural, social, psychological, environ- hybrid models were compared for their convergence speed and
mental and political factors. The results indicated a successful prediction accuracy over traditional ANN models. The resu-
prediction rate of PSO-ANN of up to 80%. Furthermore, the lts showed that swarm intelligence has higher convergence speed
technique is capable of providing faster and more accurate re- and accuracy over ANN trained with gradient descent.Lee et al.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

(2016) presented a hybrid model for estimating the quantity and slow in computing taking long time to run.
cost of waste in the early stage of construction. The approach
used ACO algorithm to optimise the selection of ANN parame-
ters. The proposed model can be used to address the cost over-
runs and improve construction waste management. The hybrid ANNs have been recognised to be more powerful than tradition-
model predicted more effectively the amount of waste concrete al mathematical and statistical methods in events of complex
in early project stage. The comparison of prediction results of qualitative and quantitative reasoning. They have been success-
ANN and ANN-ACO showed that the hybrid model had mini- fully employed in solving numerous complex nonlinear problems
mum error demonstrating a higher accuracy than ANN. of prediction, estimating, decision making, optimization, clas-
sification and selection in the fields of construction engineering
3DISCUSSION OF FINDINGS and management. They are identified to have the potentials of
dealing with noisy data and achieving high accuracy and reliable
ANN has gained considerable application to solve complex non- prediction and forecasting. Neural Networks have also been in-
linear problems in construction engineering and management tegrated with several soft computing paradigms such as fuzzy
over the few decades. Their employment of ANN in construc- logic, Case Based Reasoning, Particle Swarm Optimization,
tion cost prediction, schedule estimating, productivity forecast, Ant colony Optimization, Artificial Bee Colony and Firefly with
prediction of dispute occurrence and resolution outcomes and the view to improving accuracy, over-fitting and under-fitting of
contract performance demonstrates its potentials and robust- data and convergence speed.
ness in addressing problems that proved difficult for traditional
mathematical and statistical approaches to solve. Fundamental-
ly, the performance of ANN is data dependent which signifies REFERENCES
the importance of quality and quantity of data for training the
networks which is key to the outcomes of predictions, recogni- AbouRizk, S., Knowles, P., and Hermann, U. R. (2001). “Esti-
tion and classification as indicated byHegazy and Ayed (1998) mating labor production rates for industrial construction ac-
andWaziri and Bala (2011). tivities.” Journal of Construction Engineering and Manage-
Prediction accuracy is one key attributes of ANN over tra- ment, 127(6), 502–511.
ditional methods which adds to its popularity and usage. Ac- Adeli, H. (2001). “Neural networks in civil engineering: 1989–
curacy levels of most ANN based applications in prediction and 2000.” Computer-Aided Civil and Infrastructure Engineering,
forecasting can get to as high as 98% over test samples as evident 16(2), 126–142.
inGeiger et al. (1998),Alqahtani and Whyte (2013) andBala Adeli, H. and Karim, A. (1997). “Scheduling and cost optimiza-
et al. (2014). These performances are recorded on the bases of tion and neural dynamics model for construction.” Journal of
minimum errors over test samples as measured by MSE, MAE, Construction Engineering and Management, 123(4), 450–458.
MAPE and SSE. Findings also show that in most prediction Ahiaga-Dagbui, D. D. and Smith, S. D. (2012). “Neural netw-
problems, ANN usually demonstrates high degree of data fitting orks for modelling the final target cost of water projects.” Pro-
with high correlation coefficient usually higher than 90% (Kim ceeding of the 28th Annual ARCOM Conference, Edinburgh,
et al. 2005). United Kingdom.
Neural network models are suitable for parametric modelling Aibinu, A. (2011). “Use of artificial intelligence to predict the ac-
and most a times are used as alternative to classical modelling curacy of pre-tender building cost estimate.” Proceeding of the
techniques especially for dataset involving nonlinear relationship Conference of Management and Innovation for a Sustainable
and formed the basis for decision support tools using supervised Built Environment, Amsterdam, the Netherlands.
learning algorithm for optimal decision making which is an im- Al-Sobiei, O. S., Arditi, D., and Polat, G. (2005). “Predicting
portant activity in construction engineering and management. the risk of contractor default in saudi arabia utilizing artificial
Its performance in such applications are outstanding and com- neural network and genetic algorithm techniques.” Construc-
pares favourably with other parametric models such as regres- tion Management and Economics, 23(4), 423–430.
sion analysis (Sonmez 2004;Sonmez and Ontepeli 2009). It is Al-Tabtabai, H., Alex, P., and Tantash, M. (1998). “Prelim-
considered best for short term forecast and has the potentials inary cost estimating of highway construction using neural
for mapping uncertainties in learning. network.” Cost Engineering, 41(3), 19–24.
Despite its robustness and extremely high advantages for para- Al-Zwainy, F. M. S., Rasheed, H. A., and Ibraheem, H. F.
metric modelling and other decision making applications, ANN (2012). “Development of the construction productivity estima-
has been observed to lack general procedure especially for the se- tion model using artificial neural networks for finishing works
lection of its initial weights and other initial parameters for effec- for floors with marble.” ARPN Journal of Engineering and
tive application. It is also observed to be unsuitable for long term Applied Sciences, 7(6), 714–722.
forecasting especially for changing trends (Kim et al. 2004). To Aleksander, I. and Morton, H. (1993). Neurons and Symbols:
address this deficiency, several other learning algorithms and op- The Stuff that Mind is Made of. Chapman & Hall, London,
timization tools are been employed to develop hybrid models for United Kingdom.
improved performance. Findings revealed the successful integra- Alex, D. P., Al Hussein, M., Bouferguene, A., and Fernando,
tion of GA (Feng and Li 2013), ACO (Kayarvizhy et al. 2014), S. (2009). “Artificial Neural Network model for cost estima-
PSO (Shi and Li 2010) and fuzzy logic (Cheng et al. 2009a) with tion: City of Edmonton’s water and sewer installation ser-
ANN demonstrating favourable performance as compared with vices.” Journal of Construction Engineering and Management,
simple neural network algorithms except for ANN-GA showing 136(7), 745–756.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

Alqahtani, A. and Whyte, A. (2013). “Artificial Neural Netw- Chou, J. S. (2012). “Comparison of multilabel classification mod-
orks incorporating cost significant items towards enhancing els to forecast project dispute resolutions.” Expert Systems
estimation for (life-cycle) costing of construction projects.” with Applications, 39(11), 10202–10211.
Construction Economics and Building, 13(3), 51–64. Chua, D. K. H., Kog, Y. C., Loh, P. K., and Jaselskis, E. J.
Apanaviciene, R. and Juodis, A. (2005). “Modelling of construc- (1997). “Model for construction budget performance-neural
tion project management effectiveness by applying neural net- network approach.” Journal of Construction Engineering and
works.” Symposium on Sustainability and Value Through Con- Management, 123(3), 214–222.
struction Procurement, Salford, United Kingdom. El-Sawah, H. and Moselhi, O. (2014). “Comparative study in
Arafa, M. and Alqedra, M. (2011). “Early stage cost estimation the use of neural networks for order of magnitude cost esti-
of buildings construction projects using artificial neural net- mating in construction.” Journal of Information Technology
works.” Journal of Artificial Intelligence, 4(1), 63–75. in Construction, 19, 462–473.
Arbib, M. A. (1995). The Handbook of Brain Theory and Neural Elhag, T. M. S. and Boussabaine, A. H. (1998). “An artificial
Networks. MIT press, Cambridge, United Kingdom. neural system for cost estimation of construction projects.”
Assaf, S. A., Bubshait, A. A., Atiyah, S., and Al-Shahri, M. Proceedings of the 14th ARCOM Annual Conference, Berk-
(2001). “The management of construction company overhead shire, United Kingdom.
costs.” International Journal of Project Management, 19(5), Elhassan, S. M. M., Zawawi, N. A. W., and Ghazali, Z. B. (2012).
295–303. “Decision making framework for optimizing construction man-
Aswed, G. K. (2016). “Productivity estimating model for brick- agement objectives: A review.” Recent Advances in Engineer-
layers in construction projects using artificial neural netw- ing, 1(1), 169–173.
orks.” Al-qadisiya Journal of Engineering Sciences, 9(2), 183– ElSawy, I., Hosny, H., and Razek, M. A. (2011). “A neural net-
199. work model for construction projects site overhead cost esti-
Bala, K., Bustani, S., and Waziri, B. (2014). “A computer-based mating in egypt.” International Journal of Computers Science,
cost prediction model for institutional building projects in 8(3), 273–283.
Nigeria: An artificial neural networks approach.” Journal of Emsley, M. W., Lowe, D. J., Duff, A. R., Harding, A., and
Engineering, Design and Technology, 12(4), 518–529. Hickson, A. (2002). “Data modelling and the application of a
Bala, K. and Waziri, B. (2012). “Parametric model for pre-design neural network approach to the prediction of total construc-
building cost estimating.” Global Journal of Engineering and tion costs.” Construction Management and Economics, 20(6),
Technology, 4(1), 115–122. 465–472.
Bhokha, S. and Ogunlana, S. (1999). “Application of artificial Fang, C. F. and Froese, T. (1999). “Cost estimation of High Per-
neural networks for the forecast of building cost at the pre- formance Concrete (HPC) high-rise commercial buildings by
design stage.” Seventh East-Asia Pacific Conference on Struc- neural networks.” Available at: <>.
tural Engineering and Construction, Kochi, Japan. Fatima, A., Sekhar, T. S., and Hussain, S. A. M. (2014). “Anal-
Bouabaz, M. and Hamami, M. (2008). “A cost estimation model ysis of construction dispute resolution process using artifi-
for repair bridges based on artificial neural networks.” Amer- cial neural networks.” International Journal of Innovative Re-
ican Journal of Applied Sciences, 5(4), 334–339. search and Developments, 3(7), 81–86.
Boussabaine, A. H. (1995). “A neural network system for produc- Feng, G. L. and Li, L. (2013). “Application of genetic algorithm
tivity forecasting.” Proceedings of the 12th ISARC, Warsaw, and neural network in construction cost estimate.” Advanced
Poland. Materials Research, 756–759, 3194–3198.
Boussabaine, A. H. and Kaka, A. P. (1998). “A neural networks Galkin, I. (2002). “Crash introduction to artificial neural net-
approach for cost flow forecasting.” Construction Management works.” Available at: <>.
and Economics, 16(4), 471–479. Geiger, M., Knoblach, J., and Backes, F. (1998). “Cost esti-
Chao, L. C. and Skibniewski, M. J. (1994). “Estimating construc- mation of large scale production of sheet metal.” Production
tion productivity: neural-network-based approach.” Journal Engineering, 2(2), 81–84.
of Computing in Civil Engineering, 8(2), 234–251. Geon, N. C. (2005). “Using artificial neural networks to predict
Chau, K. W. (2007). “Application of a pso-based neural network power plant turbine hall key cost drivers.” M.S. thesis, Uni-
in analysis of outcomes of construction claims.” Automation versity of Technology Malaysia, Johor Bahru, Malaysia.
in Construction, 16(5), 642–646. Golizadeh, H., Sadeghifam, A. N., Aadal, H., and Majid, M.
Cheng, M. Y., Tsai, H. C., and Hsieh, W. S. (2009a). “Web- Z. A. (2016). “Automated tool for predicting duration of con-
based conceptual cost estimates for construction projects us- struction activities in tropical countries.” KSCE Journal of
ing evolutionary fuzzy neural inference model.” Automation Civil Engineering, 20(1), 12–22.
in Construction, 18(2), 164–172. Golpayegani, S. A. H. and Emamizadeh, B. (2007). “Designing
Cheng, M. Y., Tsai, H. C., and Sudjono, E. (2009b). “Evolution- work breakdown structures using modular neural networks.”
ary fuzzy hybrid neural network for conceptual cost estimates Decision Support Systems, 44(1), 202–222.
in construction projects.” 26th International Symposium on Graham, L. D., Forbes, D. R., and Smith, S. D. (2006). “Mod-
Automation and Robotics in construction, Austin, United S- eling the ready mixed concrete delivery system with neural
tates. networks.” Automation in Construction, 15(5), 656–663.
Chew, M. Y. L. and Tan, S. S. (2003). “Neural network ap- Günaydın, H. M. and Doğan, S. Z. (2004). “A neural network ap-
proach for grading of maintainability of wet areas in high- proach for early cost estimation of structural systems of build-
rise buildings.” International Journal of Architectural Science, ings.” International Journal of Project Management, 22(7),
4(4), 150–167. 595–602.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

Hegazy, T. and Ayed, A. (1998). “Neural network model for para- 791–797.
metric cost estimation of highway projects.” Journal of Con- Lou, Z., Gunaratne, M., Lu, J. J., and Dietrich, B. (2001). “Ap-
struction Engineering and Management, 124(3), 210–218. plication of neural network model to forecast short-term pave-
Hegazy, T., Fazio, P., and Moselhi, O. (1994). “Developing prac- ment crack condition: Florida case study.” Journal of Infras-
tical neural network applications using back-propagation.” tructure Systems, 7(4), 166–171.
Computer-Aided Civil and Infrastructure Engineering, 9(2), Lyne, R. C. and Maximinio, O. J. (2014). “An artificial neural
145–159. networks approach to structural cost estimation of building
Hong, Y., Liao, H., and Jiang, Y. (2014). “Construction engineer- projects in Palestine.” DLSU Research Congress 2014, Mani-
ing cost evaluation model and application based on RS-IPSO- la, Philippines.
BP neural network.” Journal of Computers, 9(4), 1020–1026. Maghrebi, M., Sammut, C., and Waller, T. S. (2014). “Predicting
Iranmanesh, S. H. and Zarezadeh, M. (2008). “Application of the duration of concrete operations via artificial neural net-
artificial neural networks to forecast actual cost of a project to works and by focusing on supply chain parameters.” Building
improve earned value management system.” International Research Journal, 61(1), 1–14.
Journal of Social, Behavioral, Educational, Economic, Busi- Maria-Sanchez, P. (2004). “Estimation of environmental risks in
ness and Industrial Engineering, 2(6), 658–661. construction projects in Puebla (Mexico): a neural network
Jain, M. and Pathak, K. K. (2014). “Applications of artificial approach.” WIT Transactions on Ecology and the Environ-
neural networks in construction engineering and management: ment, 77.
A review.” International Journal of Engineering Technology, Masri, S. F., Nakamura, M., Chassiakos, A. G., and Caugh-
Management and Applied Sciences, 2(3), 134–142. ey, T. K. (1996). “Neural network approach to detection of
Jamil, M., Zain, M. F., and Basri, H. B. (2009). “Neural network changes in structural parameters.” Journal of Engineering Me-
simulator model for optimization in high performance concrete chanics, 122(4), 350–360.
mix design.” European Journal of Scientific Research, 34(1), McKim, R. A. (1993). “Neural networks and identification and
61–68. estimation of risk.” AACE International Transactions, 5.1–
Kamarthi, S. V., Sanvido, V. E., and Kumara, S. R. T. (1992). 5.10.
“Neuroform-neural network system for vertical formwork sele- Morcous, G. (2002). “Comparing the use of artificial neural net-
ction.” Journal of Computing in Civil Engineering, 6(2), 178– works and case-based reasoning in modeling bridge deterio-
199. ration.” Annual Conference of the Canadian Society for Civil
Kayarvizhy, N., Kanmani, S., and Uthariaraj, R. V. (2014). Engineering, Montreal, Canada.
“ANN models optimized using swarm intelligence algorithms.” Murtaza, M. B. and Fisher, D. J. (1994). “A neural network
WSEAS Transactions on Computers, 13, 501–519. model for decision making with application in construction
Khalafallah, A. (2008). “Neural network based model for pre- management.” Journal of International Information Manage-
dicting housing market performance.” Tsinghua Science and ment, 3(2), 27–38.
Technology, 13(1), 325–328. Odeyinka, H., Lowe, J., and Kaka, A. (2002). “A construction
Kim, G. H., An, S. H., and Kang, K. I. (2004). “Comparison of cost flow risk assessment model.” ARCOM 18th Annual Con-
construction cost estimating models based on regression anal- ference, Northumbria, United Kingdom.
ysis, neural networks, and case-based reasoning.” Building and Paliwal, M. and Kumar, U. A. (2011). “Assessing the contribu-
Environment, 39(10), 1235–1242. tion of variables in feed forward neural network.” Applied Soft
Kim, G. H., Seo, D. S., and Kang, K. I. (2005). “Hybrid mod- Computing, 11(4), 3690–3696.
els of neural networks and genetic algorithms for predicting Pathak, K. K. and Agarwal, R. (2003). “Cost prediction of over-
preliminary cost estimates.” Journal of Computing in Civil head water tanks using artificial neural networks.” Journal
Engineering, 19(2), 208–211. of the Institution of Engineers. India. Civil Engineering Divi-
Lam, K. C., Hu, T., Thomas Ng, S., Skitmore, M., and Cheung, sion, 84, 153–158.
S. (2001). “A fuzzy neural network approach for contractor Pearce, A. R. (1997). “Cost-based risk prediction and identifica-
prequalification.” Construction Management and Economics, tion of project cost drivers using artificial neural networks.”
19(2), 175–188. Final report submitted to the Graduate Program, Georgia In-
Lee, D., Kim, S., and Kim, S. (2016). “Development of hy- brid stitute of Technology, Atlanta, United States.
model for estimating construction waste for multifamily Petruseva, S., Zileska-Pancovska, V., and Zujo, V. (2013). “Pre-
residential buildings using artificial neural networks and ant dicting construction project duration with support vector ma-
colony optimization.” Sustainability, 8(9), 870–883. chine.” International Journal of Research in Engineering and
Lhee, S. C., Flood, I., and Issa, R. R. (2014). “Development of Technology, 2(11), 12–24.
a two-step neural network-based model to predict construc- Pewdum, W., Rujirayanyong, T., and Sooksatra, V. (2009).
tion cost contingency.” Journal of Information Technology in “Forecasting final budget and duration of highway construc-
Construction, 19(24), 399–411. tion projects.” Engineering, Construction and Architectural
Li, H. and Love, P. E. D. (1999). “Combining rule-based ex- Management, 16(6), 544–557.
pert systems and artificial neural networks for mark-up es- Polat, G. (2012). “ANN approach to determine cost contingen-
timation.” Construction Management and Economics, 17(2), cy in international construction project.” Journal of Applied
169–176. Management and Investments, 1(2), 195–201.
Liu, J. B. and Guo, F. (2014). “Construction quality risk man- Portas, J. and AbouRizk, S. (1997). “Neural network model for
agement of projects on the basis of rough set and neural net- estimating construction productivity.” Journal of Construc-
work.” Computer Modelling and New Technologies, 18(11), tion Engineering and Management, 123(4), 399–410.

Waziri et al./International Journal of Architecture, Engineering and Construction 6 (2017)50-60

Rudomin, P., Arbib, M. A., Cervantes-Perez, F., and Romo, neural network in the cost estimation of highway engineering.”
R. (1993). Neuroscience: From Neural Networks to Artificial Journal of Computers, 5(11), 1762–1766.
Intelligence. Springer, Berlin, German. Wang, Y. M. and Elhag, T. M. (2007). “A comparison of neural
Rumelhart, D. E. (1986). “Learning internal representations by network, evidential reasoning and multiple regression analysis
error propagation.” Parallel Distributed Processing, 1, 318– in modelling bridge risks.” Expert Systems with Applications,
362. 32(2), 336–348.
Savin, D., Alkass, S., and Fazio, P. (1998). “Calculating weight Waziri, B. S. (2010). “An artificial neural networks model for pre-
matrix of neural network for resource leveling.” Journal of dicting the construction cost of institutional building projects
Computing in Civil Engineering, 12(4), 241–248. in Nigeri.” Proceedings of 2nd West Africa Built Environment
Sawhney, A. and Mund, A. (2001). “Intellicranes: An integrated Research Conference, Accra, Ghana.
crane type and model selection system.” Construction Man- Waziri, B. S. (2012). “Modelling the performance of tradition-
agement and Economics, 19(2), 227–237. al contract projects in Nigeria: An artificial neural networks
Setyawati, B. R., Sahirman, S., and Creese, R. C. (2002). “Neural approach.” Proceedings of 4th West Africa Built Environment
networks for cost estimation.” AACE International Transac- Research Conference, Abuja, Nigeria.
tions, ES13.1–ES13.9. Waziri, B. S. and Bala, K. (2011). “Predicting construction cost
Sharmik, S., Lekha, G., and Kanshik, S. (2016). “Cost and time of public building projects using regression and neural net-
effective prediction of soil characteristics using artificial neu- work models.” International Journal of Architecture and Built
ral networks model.” International Journal of Innovative Re- Environment, 3(1), 85–91.
search in Science, Engineering and Technology, 5(3), 3829– Williams, T. P. (1994). “Predicting changes in construction cost
3834. indexes using neural networks.” Journal of Construction En-
Shi, H. W. and Li, W. Q. (2010). “Integration of fuzzy logic, gineering and Management, 120(2), 306–320.
particle swarm optimization and neural networks in quality Wilmot, C. G. and Mei, B. (2005). “Neural network modeling
assessment of construction project..” Journal of Software En- of highway construction costs.” Journal of Construction En-
gineering, 5(7), 737–744. gineering and Management, 131(7), 765–771.
Shi, J. J. (1999). “A neural network based system for predict- Xiang, P. C. and Luo, K. (2012). “The evaluation for the be-
ing earthmoving production.” Construction Management and havioural risk of participants in construction projects based
Economics, 17(4), 463–471. on back propagation neural network.” Advances in Informa-
Shtub, A. and Versano, R. (1999). “Estimating the cost of tion Sciences and Service Sciences, 4(14), 97–107.
steel pipe bending, a comparison between neural networks and Yadav, R., Vyas, M., Vyas, V., and Agrawal, S. (2016). “De-
regression analysis.” International Journal of Production Eco- velopment of cost estimating method for bricklayer cost.” In-
nomics, 62(3), 201–207. ternational Journal of Engineering Research and Technology,
Sivanandam, S. N. and Deepa, S. N. (2006). Introduction to Neu- 5(1), 430–432.
ral Networks using Matlab 6.0. Tata McGraw-Hill Education, Yahia, H., Hosny, H., and Razik, M. E. A. (2011). “Time contin-
Columbus, United States. gency assessment in construction projects in Egypt using arti-
Sodikov, J. (2005). “Cost estimation of highway projects in de- ficial neural networks model.” International Journal of Com-
veloping countries: artificial neural networks approach.” Jour- puter Science, 8(4), 523–531.
nal of the Eastern Asia Society for Transportation Studies, 6, Yeh, I. C. (1998). “Modeling of strength of high-performance con-
1036–1047. crete using artificial neural networks.” Cement and Concrete
Sonmez, R. (2004). “Conceptual cost estimation of building Research, 28(12), 1797–1808.
projects with regression analysis and neural networks.” Cana- Yitmen, I. and Soujeri, E. (2010). “An artificial neural network
dian Journal of Civil Engineering, 31(4), 677–683. model for estimating the influence of change orders on project
Sonmez, R. and Ontepeli, B. (2009). “Predesign cost estimation performance and dispute resolution.” Proceedings of the In-
of urban railway projects with parametric modeling.” Journal ternational Conference on Computing in Civil and Building
of Civil Engineering and Management, 15(4), 405–409. Engineering, Nottingham, United Kingdom.
Vahdani, B., Mousavi, S. M., Mousakhani, M., Sharifi, M., and Zin, M. R., Mansur, S. A., Bakri, A., and Caren, T. C. L. (2006).
Hashemi, H. (2012). “A neural network model based on sup- “Predicting the performance of traditional general contract
port vector machine (SVM) for conceptual cost estimation of (TGC) projects: A neural network based approach.” Proceed-
construction projects.” Journal of Optimization in Industrial ings of the 6th Asia-Pacific Structural Engineering and Con-
Engineering, 10, 11–18. struction Conference, Kuala Lumpur, Malaysia.
Wang, X. Z., Duan, X. C., and Liu, J. Y. (2010). “Application of

View publication stats