See discussions, stats, and author profiles for this publication at: [Link]

net/publication/351144656

Application of machine learning in predicting construction project profit in

Ghana using Support Vector Regression Algorithm (SVRA)

Article in Engineering Construction & Architectural Management · April 2021

DOI: 10.1108/ECAM-08-2020-0618

CITATIONS READS

4 520

4 authors:

Emmanuel Adinyira Emmanuel Akoi-Gyebi Adjei

Kwame Nkrumah University Of Science and Technology Accra Technical University
145 PUBLICATIONS 1,033 CITATIONS 3 PUBLICATIONS 12 CITATIONS

SEE PROFILE SEE PROFILE

Kofi Agyekum Frank D. K. Fugar

Kwame Nkrumah University Of Science and Technology Kwame Nkrumah University Of Science and Technology
179 PUBLICATIONS 1,043 CITATIONS 62 PUBLICATIONS 786 CITATIONS

SEE PROFILE SEE PROFILE

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

Persons with disabilities and their accessibilty of built infrastructure facilities View project

National Building Regulation View project

All content following this page was uploaded by Frank D. K. Fugar on 26 September 2021.

The user has requested enhancement of the downloaded file.

 The current issue and full text archive of this journal is available on Emerald Insight at:
[Link]

Application of machine learning in Machine

learning and
predicting construction project construction
project profit
profit in Ghana using Support
Vector Regression 1491
Algorithm (SVRA) Received 9 August 2020
Revised 5 December 2020
24 January 2021
Emmanuel Adinyira 24 March 2021
Accepted 30 March 2021
Construction Technology and Management,
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
Emmanuel Akoi-Gyebi Adjei
Building Technology, Accra Technical University, Accra, Ghana and
Construction Technology and Management,
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
Kofi Agyekum
Department of Construction Technology and Management,
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana, and
Frank Desmond Kofi Fugar
Construction Technology and Management,
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana

Abstract
Purpose – Knowledge of the effect of various cash-flow factors on expected project profit is important to
effectively manage productivity on construction projects. This study was conducted to develop and test the
sensitivity of a Machine Learning Support Vector Regression Algorithm (SVRA) to predict construction project
profit in Ghana.
Design/methodology/approach – The study relied on data from 150 institutional projects executed within
the past five years (2014–2018) in developing the model. Eighty percent (80%) of the data from the 150 projects
was used at hyperparameter selection and final training phases of the model development and the remaining
20% for model testing. Using MATLAB for Support Vector Regression, the parameters available for tuning
were the epsilon values, the kernel scale, the box constraint and standardisations. The sensitivity index was
computed to determine the degree to which the independent variables impact the dependent variable.
Findings – The developed model’s predictions perfectly fitted the data and explained all the variability of the
response data around its mean. Average predictive accuracy of 73.66% was achieved with all the variables on
the different projects in validation. The developed SVR model was sensitive to labour and loan.
Originality/value – The developed SVRA combines variation, defective works and labour with other
financial constraints, which have been the variables used in previous studies. It will aid contractors in
predicting profit on completion at commencement and also provide information on the effect of changes to cash-
flow factors on profit.
Keywords Cash flow, Construction project, Machine learning, Profit, Productivity, Support vector regression
Paper type Research paper

Introduction
Engineering, Construction and
The high level of uncertainties encountered in the construction industry worldwide renders Architectural Management
most construction companies bankrupt (Majer et al., 2020). Of all the uncertainties that Vol. 28 No. 5, 2021
pp. 1491-1514
contribute to the bankruptcy of construction companies, financial and budgetary factors play © Emerald Publishing Limited
0969-9988
significant roles (Majer et al., 2020; Arditi et al., 2000). It is established that over 60% of DOI 10.1108/ECAM-08-2020-0618
ECAM construction company failures worldwide have their roots traced to financial issues, which
28,5 affect their profitability status (Majer et al., 2020). According to Zhu et al. (2019), most of these
construction projects fail to achieve their expected profits because of the difficulties in
managing certain risks in a more complex web of project organisation. Profitability is,
therefore classified as one of the most obvious causes of insolvency in the construction
industry (Arditi et al., 2000).
Profitability essentially can be argued to mean a measure of productivity if improved
1492 productivity is viewed as an increase in value-added outputs with fewer resources (Snyman
and Smallwood, 2017). When an organisation focuses more on increasing the amount of work
it takes on without considering the profits, the organisation will often settle for smaller profit
margins or secure work that is not profitable with the purpose of increasing its volume of
work (Snyman and Smallwood, 2017). The issue of construction failure as a result of
dwindling profit has been of interest to the scientific community. Construction is a very risky
business (Lee, 2009) and operates within a competitive environment that usually compels
contractors to introduce lower profit margins in their tender for projects to win bids.
Unfortunately, the nature of construction projects that is largely characterised by constant
environmental changes, pressures to maintain schedules and to reduce costs, among other
things, creates a significant challenge to managing projects (Cheng et al., 2015). As a matter of
importance, there is a need to put in place measures that can provide construction companies
with accurate information on critical factors like possible profit status (Zhu et al., 2019).
Unfortunately, this is not the case because, in practice, contractors are usually unwilling to
conduct further profit performance predictions due to time and cash restrictions (Zhu et al.,
2019). Of these two factors, as revealed by Zhu et al. (2019), cash has been identified as a
critical factor that imposes a significant influence on project profitability (Jiang et al., 2011).
Currently, in Ghana, because construction companies operate in highly competitive
environments and cannot survive without effective management, they have resorted to the
introduction of low-profit margins in tender bids to compete within the industry (Adjei et al.,
2018). This decision by such companies in Ghana in the long term affect their liquidity, which
then leads to the failure of projects and subsequent bankruptcy of such companies (Adjei
et al., 2018). Poor cash-flow control tends to cause project failure for contractors, as a result of
the liquidity shortage for supporting their daily activities (Khosrowshahi and Kaka, 2007).
Due to this, Mohsin et al. (2014) posit that reliably predicting cash flow over various phases of
a project is highly desirable because the project manager will be in a better position to curb
the potential financial issues that can hinder the success of the project. Cash-flow forecasting,
therefore, is an important tool to evaluate the distribution of expenditure and revenues of
projects.
Several researchers have developed models that have the potential to aid contractors to
plan their cash flows (El-Kholy, 2014; Jiang et al., 2011). Authors who have developed most of
these models have employed more advanced forecasting algorithms like Kalman filters,
Artificial Neural Networks (ANN) and Support Vector Machines (SVM) (Sapankevych and
Sankar, 2009). Unfortunately, Mohsin et al. (2014) reckon that these models are either project-
specific or do not address the consequences of failure to meet the minimum funds required for
cash outflow. This notwithstanding, Cheng et al. (2015) has been of the view that since SVM
has been utilised in several applications in construction engineering, it can be a promising
tool for cash-flow problems.
This study has become necessary because apart from productivity, two basic threats that
affect the financial future of construction organisations are profitability and cash flow
(Peterson, 2009). With the issue of productivity and cash flow well reported in the
international literature, there is the need to focus on how contractors within a typical
developing country setting like Ghana can set up an appropriate decision support system
that can guide construction companies (especially during the bidding process) to maximise
their profits. Since the impact of cash flow on the profitability of contractors is well Machine
established, this study implores cash-flow factors to develop a model that can predict profit learning and
on construction projects in Ghana through Support Vector Regression Algorithm (SVRA).
One of the key contributions of this study stems from the fact that the SVRA model developed
construction
contributes its quota to the state-of-the-art by potentially filling the profit measurement gap project profit
with the incorporation of variation, defective works and labour to financial constraints which
have been the variables used in previous studies. The study begins with an introduction to
the central theme of the study, followed by a critical review of related literature. The research 1493
methodology is then presented, after which the results are presented and discussed, and then
finally, a conclusion is drawn.

Literature review
This section presents a comparative review of literature related to the subject under
investigation. It begins with a general review of literature on the construction industry and
profit challenges, and then narrows down to cash flow as a profit challenge and discusses the
factors that affect it. The section continues with a review of literature on various theories of
profits and then proposes a conceptual framework that relates the discussed theories of profit
to construction project profit. The section concludes with a review of various predictive
models, with emphasis laid on Support Vector Regression in Machine Learning.

The construction industry and profit challenges

Akintoye and Skitmore (1991) defined profitability as a percentage of the profit of turnover
(POT) or returns on capital investment (ROI). It is a residual of sales revenue once all costs
(including interest payment on debt) have been deducted (Achenef, 2016). Malik (2011)
classified profitability as being one of the most important objectives of financial management
because a business that is not profitable cannot survive. Wright (1977) considered
profitability as a function of three factors, that is sales volume (work done), the capital
investment needed to support business and the margin of profit earned. The level of
profitability of any construction company depends on the demand for building works from
both the private and public sectors (Lee, 2009). This means that a reduction in turnover
generally reduces the level of profitability (Lee, 2009). Key challenges confronting the
profitability of construction firms include the size of the company (Akintola and Skitmore,
1991); the extent of mechanisation (Hillebrandt, 2000); the extent of subcontracting (Reich,
2000); the extent of labour-only subcontracting (Langford and Chan, 1987); the volume of
sales and profit margins (Hillbrandt et al., 1995); gearing factor and turnover of capital
employed (Wright, 1977); property and construction cycles (Hillbrandt, 2000); competition
(Akintoye and Skitmore, 1991) and macro-economic factors (Lee, 2009). A few of these
challenges are discussed in the further paragraphs.
Size of the company. Though there are mixed reactions regarding the influence of firm size
on profitability, studies related to the construction industry have confirmed that the size of a
construction company has a positive correlation with its profit ratio to an optimum level
(Akintola and Skitmore, 1991). Hillbrandt (2000) was of the view that this assertion may be
true because larger construction companies had the potential resources to heavily invest in
construction activities, unlike the smaller ones. Spedding (1977) was also of the view that such
large companies are more efficient and well organised in their managerial strategies
compared to the smaller ones. Lee (2009) supported these assertions and added that when it
comes to estimating, pricing and production, larger companies are very consistent. Though
there is evidence in the literature that suggests that large contractors are consistent in
maintaining their profitability levels as compared to smaller ones, the volume of size has yet
to be analysed (Lee, 2009).
ECAM Competition. The construction industry is very competitive by its nature. The growing
28,5 importance of competitiveness in the industry depicts it as a predictor of business
performance (Oyewobi et al., 2014). The competitive environment necessitates that
construction organisations are more strategic and proactive to increase their chances of
survival (Oyewobi et al., 2014). According to the standard economic theory, increased
competition among firms results in lower profitability (Lee, 2009). The competitiveness of the
construction industry motivates contractors to bid for jobs using low mark-ups to provide a
1494 high chance of job acquisition (Halim, 2014). When this happens, firms with insufficient profit
margins may go bankrupt, thereby resulting in their tendencies to collapse. A similar
scenario was reported by Mahamid (2011), who indicated that in the USA, insufficient profit
caused about 27% of construction firms to fail. It must be noted that the unfamiliarity of the
risk involved in this competition that normally occurs during bidding in the construction
industry has contributed to low profitability levels in the industry (Yean et al., 2005).
Gearing factor and turnover of capital employed. The gearing factor measures the quantum
of investment made against the volume of sales or work done (Wright, 1977). The gearing
ratio is an important measure of the stability of a company since it is considered when raising
external capital (Tunji et al., 2015). If a company is already highly geared, it might find it
extremely difficult to raise additional fund as the would-be lender may take a closer look at its
structure and believe that the company might not be able to settle the debts as at when due
(Tunji et al., 2015). This is because it may already be exposed to so many creditors. The effect
of having excess gearing is that, such a company would have to accumulate the higher
amount of profit before interest and tax to be able to meet the demand for interest payment.
Improving the effectiveness with which firms use their capital employed can improve the
profitability of the firm (Lee, 2009). The turnover of capital employed in any construction
business is largely determined by how that firm invests its capital in fixed assets, among
other things (Lee, 2009). As a matter of fact, the lesser the investment expenditure made in, for
instance, fixed assets to support a given level of activity, the more profitable that firm will be
(Lee, 2009). When a construction company exhibits a high gearing ratio for loans, its
profitability and chances of survival in the long term are greatly impacted (Lee, 2009).

Cash flow as a profit challenge and factors affecting it

Though there is a close relationship between cash flow and profit, the two are not the same.
Cash flow represents the cash inflows and outflows from the business. Subtracting cash
outflows from the cash inflows results in the net cash flow (Hofstrand, 2013). Profitability, on
the other hand, represents the income and expenses of a business (Hofstrand, 2013).
Subtracting the expenses from the income either results in a profit or a loss. Despite this
difference, making a profit generates cash flow (Tracy and Tracy, 2015). This, therefore,
makes it imperative for any company to have sufficient cash to pay its creditors, suppliers,
employees, etc. (Omopariola et al., 2020). According to Pansegrauw (2019), whilst revenue is
vanity, cash flow is sanity and cash is king, and it constitutes an essential construction
company resource. Omopariola et al. (2020) stressed that the significance of cash flow relates
to cash as the basic source of every successful construction project. As a matter of fact, a
stable financial performance that is achieved through a productive cash-flow analysis assists
construction organisations to exploit their investment opportunities (Seo et al., 2018). The
contractor sees cash flow as very important because it reflects a project’s financial
performance prior to the completion of the contract and the settling of final accounts (Usman
et al., 2016).
Construction is undertaken within a risky environment and subjected to significant
uncertainties that include the need for capital, delays in client payments and varying interest
rates between the contract end time and final payment (Barbosa and Pimentel, 2010).
During high-interest rates and inflation levels periods, cash-flow forecasting becomes more Machine
important, and this makes it an ideal tool to evaluate the distribution of expenditure and learning and
revenues for projects concerning project time (Mohsin et al., 2014). The non-availability or
inadequacy of a structure to manage cash-flow results in liquidity problems that affect
construction
working capital without warning, and this defies the sustainability of projects. project profit
Numerous factors have been identified from the literature to have an impact on cash flow
(Olatunji, 2010). Studies such as those of Buertey and Adjei-Kumi (2012), Liu et al. (2009) and
Odeyinka et al. (2013), have particularly identified several factors that impact the cash flow of 1495
construction activities from the perspective of developing countries. Table 1 presents cash
flow factors gleaned from the extensive review of related literature. For lack of space, only a
few of the critical factors that affect the cash flow of projects have been discussed. Other
equally critical ones are summarised in Table 1.
In a study by Asante (2014), it was revealed that most Ghanaian contractors are not
sufficiently capitally equipped, and this influences their loan acquisition possibilities in
addition to the contractor’s ability to deliver projects successfully. Contractors thus
extremely depend upon outsourced capital from suppliers, subcontractors and advance
payments from clients (Asante, 2014). It becomes increasingly difficult under such a situation
to persuade creditors and potential lenders of the short-term inadequacy of cash. The unusual
remedy is, therefore, to issue a loan at a high capital cost to cover the risk posed (Gundecha,
2013). This increase in loan cost affects cash flow on projects negatively and further causes
greater operating cost that reduces profit (Bolek and Wili
nski, 2012). Contractors, thus, strive
to take the heavy daily construction expenses that often comprise huge sums of money

Sn Cash flow factors Sources

1 Progress payment duration RICS (2011)

2 Progress payment conditions RICS (2011)
3 Advance payment Asante (2014), Su and Lucko (2015)
4 Percentage of retention Kaka and Cheetham (1997), Hughes et al. (2000), Park et al.
(2005), Liu et al. (2009)
5 Time of releasing the retention Park et al. (2005), Liu et al. (2009)
6 Limit of retention RICS (2011)
7 Repayment of loan Bolek and Wili
nski (2012)
8 Withholding tax Buertey and Adjei-Kumi (2012)
9 Payment of creditors Asante (2014)
10 Overwork measurement Liu et al. (2009), Buertey and Adjei-Kumi (2012), Gundecha
(2013), Aziz (2013)
11 Under work measurement Liu et al. (2009), Buertey and Adjei-Kumi (2012)
12 Material cost Park et al. (2005)
13 Wages of labour and staff Park et al. (2005)
14 Plant and equipment cost Park et al. (2005)
15 Bank interest rate Barbosa and Pimentel (2010)
16 Sub-contracting Lee (2009)
17 Delay of making payments to suppliers Harris and McCaffer (2005)
18 Price variation Harris and McCaffer (2005), Gundecha (2013)
19 Work execution errors Long (2015)
20 Replacement of defective work Hughes et al. (2000)
21 Variation of works Mohsin et al. (2014)
22 Claims Sepp€al€a (2005), Long (2015), Usman et al. (2016)
23 Phasing of projects RICS (2011)
24 Overheads Lee (2009) Table 1.
25 Number of projects being executed by a Buertey and Adjei-Kumi (2012) Construction project
contractor cash flow factors
ECAM coupled with payment delay (Sambasivan and Soon, 2007; El-Kholy, 2014). This problem has
28,5 been identified in other studies that iterated that a toxic blend of cash-flow reliance on bank
loans, high-interest rates paid, and cash-flow mismanagement is the prime source of business
failure in developing countries (Enshassi et al., 2009; Barbosa and Pimentel, 2010; Jiang
et al., 2011).
Duration has also been identified as one of the requirements on which projects are
awarded. Olatunji (2010) defined project duration as the time taken to execute the project
1496 tasks from the inception of the site to the delivery of the project. Because duration plays a
major role in the cash flow of a project, studies have concluded that timeliness that regulate
the completion of the contract on the scheduled date are measures of contractors’ quality
performance that eliminates or minimises delays (Yasamis et al., 2002; Olatunji, 2010). Cost
elements relating to labour, materials and equipment constituting the greater percentage of
project costs are assumed to be a fixed proportion of the total cost over the project’s duration
(Park et al., 2005). However, changes in the duration of contracts occur because of a variation
order or change of contract conditions, and this should be accompanied by an adjustment of
weights of cost categories. Duration is, therefore considered to be an important factor that
affects the cash flow of construction companies (Haseeb et al., 2011; Owalabi et al., 2014).
In addition to the duration of projects, the margin of profit has also been identified to play a
critical role in the determination of the cash flow of any project. Its role is critical because it
gauges the financial strength of a business (Mahamid, 2011). Avetisyan et al. (2020) defined
profit of margin as the ratio of profit earned to sales receipts (costs) over a period. This means
that the profit margin measures the amount of profit that a firm accrues from sales of a
product or service. There is a tendency for less capital to be locked up in a contract when the
margin of profit is higher (Avetisyan et al., 2020). On the other hand, a greater amount of
capital is locked up when the margin of profit is less (Mahamid, 2011). With the construction
industry being associated with low-profit margins due to competition, contractors are
encouraged to rely on cash flow as a mechanism to generate profit. A company must,
therefore, operate efficiently to recover not only the cost of production but to provide
compensation for clients in exchange for risk acceptance to generate a sizeable profit margin
(Avetisyan et al., 2020; Asante, 2014).
Payment delay is also recognised as a serious challenge confronting the construction
industry of many countries (Ramachandra and Rotimi, 2011). This issue consequently exerts
a financial burden on contractors who may not have large capital as the clients hence impacts
on contractors’ cash flow negatively in meeting the financial commitment to complete
scheduled works. Available credit facilities could ease the financial burden of contractors and
cover payment delay, thereby enhancing the cash flow on projects and the company at large.
Asante (2014) reported that the most important aspect of cash-flow management is to avoid
extended cash shortages that are caused by having too great a gap between cash inflows and
outflows. He further cited empirical evidence by Jaafar and Abdul-Aziz (2005) that smaller
capital is used by contractors to start construction firms compared to other businesses. This
can be associated with the advance payment received from clients and the credit influence of
suppliers as well as the capital support given by subcontractors, although they are paid when
contractors receive payment from clients.
An equally important factor that affects the cash flow of companies is credit. Crediting is a
delayed payment arrangement through which a buyer takes possession of something and
pays later. It enhances the cash flow of contractors in the capital-intensive construction
industry, where contractors are paid after executing works successfully and certified. Harris
and McCaffer (2005) suggested that the delay of one-week is normal for labour payment while
a period of three to six-weeks is also normal for plant hirers and material suppliers. The time-
lapse indicated above is said to be a normal trading arrangement, and any period further than
this range undermines the commercial confidence in a company.
 Retention is also reported to be important when dealing with cash-flow issues. Retention is Machine
the cash retained by the employer from the contractor that protects the employer from a learning and
contractor’s insolvency as well as ensuring that the contractor finishes the work as the
contract stipulates (Hughes et al., 2000). It provides clients of the fund to depend on when
construction
contractors fail to perform due to incompetence and bankruptcy. It also motivates contractors project profit
to undertake any minor outstanding works and the repair of defects after practical
completion. Hughes et al. (2000) reported that most contractors are subjected to cash
retention, and main contractors likewise withhold part of subcontractors’ payments. Lower 1497
rates of retention are applied on large contracts since its application on the value will result in
a large sum of cash. The percentage of retention is reported to be in the range of 1–15%, with
3% as median (Hughes et al., 2000). An increase in retention rate from 2.5 to 5.0% would cause
a corresponding increase in working capital needed from 2.61 to 4.05% of annual turnover
(Kaka and Cheetham, 1997).

Theories of profit
Profit is the financial benefit realised when work is done, and also when the expenses incurred
by a business are exceeded by the amount of revenue generated by the activities of that
business (Avetisyan et al., 2020). Mathematically, profit can be expressed as (Avetisyan et al.,
2020; Halim, 2014):
sProfit ¼ total revenue  total expenses (1)

Revenue is the overall amount of income received from work done or by sales of goods or
services related to the primary operations of the company (Halim, 2014). It is one of the items
required to continuously determine, control, and plan in connection with profit. Total
expenses, on the other hand, is the total cost incurred in the construction or production of the
goods or service (Avetisyan et al., 2020).
The succeeding subsections present some of the notable theories on profit.

Modern theory of profit

This theory of profit also termed demand and supply of profit describe profit as the net
income of an entrepreneur and an entrepreneur as a business enterprise itself (Marchal, 1951).
Profit is regarded in this theory as the reward of an entrepreneur and overseen by the demand
for and supply of an entrepreneur (Marchal, 1951). Demand for entrepreneurs mostly depends
on the level of industrial development, elements of industrial uncertainty, the scale of
production, and marginal revenue productivity of entrepreneurship. The expected profit is
likely to be high when the level of industrial growth is high; the scale of production is large in
addition to an increase in efficiency and productivity (Marchal, 1951).

Rent theory of profit

Walker’s rent theory of profit considers profit as the rent of ability (Vercellone, 2008). This
theory makes a comparative study between grades of land and the entrepreneur’s abilities.
Higher ability entrepreneurs earn profits as superior land earns rent (Chendroyaperumal,
2009). Just as there is the marginal or no rent land, similarly, there exists a marginal or no
profit entrepreneur who earns only wages of management. Industries are managed and run
by marginal entrepreneurs like marginal land. A land which at its margin earns no rent so as
the marginal entrepreneur earns no profit.

Dynamic theory of profit

Clarks propounded the dynamic theory of profit and stated that profit is the difference
between the price and the cost of the commodity (Bowman, 2014). It is the result of dynamic
ECAM change. In a stationary state in which there are static economic conditions of demand and
28,5 supply, no real or pure profit are realised as a surplus. The amount of capital invested,
production methods, managerial organisation, technology and demand pattern continues to
be constant in a stationary economy (Chendroyaperumal, 2009). Price, therefore, tends to
equal average costs under competitive conditions, which results in zero surpluses. No pure
profit is earned. However, there may be some frictional profits emerging due to frictions in the
system and this cannot be regarded as real Profit (Bowman, 2014). Profits are earned because
1498 of changes or an increase in population, tastes, and wants. Also, profit is influenced by capital
formation, advancement in technology, and changes in the form of business organisation
(Chendroyaperumal, 2009).

Wage theory of profit

Taussig and Davenport’s theory explains that profits are best regarded as a basic form of
wages that accrue to the entrepreneur because of special ability (Bowman, 2014;
Chendroyaperumal, 2009). The theory reasoned that there exists a close comparison between
labourers and entrepreneurs in that, as wages are received by labourers for rendered services,
so will entrepreneurs earn profit for a service (Bowman, 2014; Chendroyaperumal, 2009).

Managerial efficiency theory of profit

This theory, also known as the compensatory theory of profits, acknowledges that some
firms are more efficient than others in terms of management of productive operations and
meeting the needs of consumers successfully (McGuigan et al., 2017). A firm with an average
level of efficiency receives an average rate of return while that with higher managerial skills
and production efficiency are required to be compensated by above-normal profits
(McGuigan et al., 2017; Chendroyaperumal, 2009; Sanyal, 2019).

Frictional theory of profit

This theory assumes that a normal proportion of profit exists, which is a return on capital
paid to the owners of capital as repayment for saving and investment of funds rather than to
spend all their income (McGuigan et al., 2017; Sanyal, 2019). In a still economy where changes
are not expected in demand or cost condition occurs, firms would be earning a normal rate of
profit on capital and entrepreneurial capacity at equilibrium over the long run (Sanyal, 2019).
Accrued economic profit under these conditions would not have ensued to a firm. The
frictional theory of profit, therefore, describes the shocks or disorders that occasionally arise
in an economy due to unexpected changes in demand for product or cost conditions which
cause an imbalance condition (McGuigan et al., 2017; Sanyal, 2019).

Monopoly theory of profit

This theory of profit is credited to the power monopoly firms enjoy (Chendroyaperumal,
2009). Products from these firms having the monopoly power are restricted and this is
accompanied by prices higher than within perfect competition (McGuigan et al., 2017; Sanyal,
2019). There exist robust barriers preventing the entry of new firms to add to firms under
monopoly. Hence, monopolistic firms earn and continue to receive economic profits even in
the long run. This power of monopoly may arise owing to sole control over some vital raw
material needed to produce a product, from economies of scale, legal sanction or ownership
patents and Government restrictions on the import of a commodity (McGuigan et al., 2017;
Sanyal, 2019).

Innovations theory of profit

This states that economic profit is generated from successful innovations introduced by
entrepreneurs (Chendroyaperumal, 2009). The main function of an entrepreneur is to present
innovations in the economy and gain profits as a reward for accomplishing the role Machine
(McGuigan et al., 2017; Sanyal, 2019). learning and
construction
Risk and Uncertainty Bearing theory of profit project profit
Risk and Uncertainty Bearing theory states that profits are the obligatory reward of an
entrepreneur for accepting risk and uncertainty in a changing economy (McGuigan et al.,
2017; Chendroyaperumal, 2009; Sanyal, 2019). Profits occur because of future uncertainty; 1499
hence, entrepreneurs undertake production works under uncertain conditions (Sanyal, 2019).
Figure 1 conceptualises project profit in the context of this study by relating the above-
discussed theories of profit to construction project profit.

Predictive models
Several studies have been undertaken for forecasting planning, and management concerning
construction and other industries. In all these studies, mathematical models have been found
to retain greater advantages of being practical, simple, fast and not requiring extensive
information about projects (Khosrowshahi and Kaka, 2007). According to Adjei et al. (2019b),
various comparative studies have been conducted between the traditional or conventional
techniques of forecasting or predicting. Whereas regression models that basically predicts
numbers abound in literature, there are also classification models that seek to predict
groupings. While traditional regression procedures derive a function f(x) that has the least
deviation between predicted and observed values for the training examples, machine
learning regression techniques enable computer programs to automatically enhance
performances of certain tasks through understanding (Pham and Afify, 2005).
Support Vector Machine Models (SVMMs) make use of machine learning regression
techniques and are known to demonstrate high predictive and accuracy performance
comparable to other regression techniques (Pham and Afify, 2005). SVMs are regressors
that perform the regression and predict continuous ordered variables using algorithms

Work execution errord, defective works, claims,

overheads, profit margin, phasing projects,
under measurement, over measurement, number
of projects, interest rate

Labour
plant
Subcontractor, suppliers, Managerial Efficiency
creditors Theory
Monopoly Theory Rent Theory
Wage Theory
Profit

Innovative Theory
Dynamic Theory Risk and Uncertainty
Theory

New Technology and methods,

source materials, organisational
Cashflow forecasting, material, change, new product, variety of
product etc
plant. labour
Cashflow forecasting, pricing strategies, price Figure1.
variation, prgress paym ent duration / Relating theories of
condtion, retention, repayment of loan,
interest rate
profit to construction
project profit
ECAM (Awad and Khanna, 2015). Advantages of Support Vector Regression (SVR) includes: the
28,5 capability to model non-linear relationships, solves regression function by selecting only the
necessary data (support vectors) resulting in a sparse solution, regression function is related
to a quadratic problem that has a typical comprehensive solution, can be employed when
there are few samples than variables called small and large problems, its computational
complexity does not depend on the dimensionality of the input space and its excellent
generalisation capability with high prediction accuracy. These advantages notwithstanding,
1500 in SVR, there is no structured method in the choice of kernel function and there is the need for
the user to describe several free parameters as the overview performance of SVR models is
dependent on the precise setting of free parameters. SVR raises a quadratic optimisation
problem of the same training data set.

Support vector regression in machine learning

The movements of the Fourth Industrial Revolution (4IR) as seen in literature can be
classified into three distinct features: defined task, undefined task and improvement
possibility to existing technologies/methods (Moon et al., 2020). Whereas 4IR in terms of
defined task, seeks to replace the human workforce, in the case of an undefined task, it is more
about diverse and complex data sets (Moon et al., 2020). It employs big data and machine
learning techniques to effectively process and use the collection of unlimited data from
numerous sources. It looks at how and what can be learned from data. Big data analytics pulls
from existing information to look for emerging patterns that can help shape decision-making
processes (Moon et al., 2020). Machine learning, on the other hand, learns from existing data
and provides the foundation required for a machine to teach itself. Big data analytics reveals
patterns through classifications and sequence analysis, but machine learning takes this
concept a step higher by using the same algorithms that big data analytics uses to
automatically learn from the collected data (Moon et al., 2020).
The basis of one of machine learning’s most popular techniques, Support Vector Machines
(SVMs), was developed by Vapnik and it is categorised under supervised learning (Vapnik,
1998). SVMs solve binary classification problems by formulating them as convex
optimisation problems (Awad and Khanna, 2015). There are fundamentally two phases in
supervised learning namely; learning and testing phases (Listiani, 2009). The learning phase
is where training data is employed to develop a mathematical model to explain the
relationship between variables. The significance of training in machine learning is to assist in
effectively adjusting and optimising the model parameters with the use of the categorised
quantitative dataset (Kolisetty and Rajput, 2020; Huang, 2007; Verma et al., 2016). On the
other hand, the test phase is when the developed model is employed to predict the results of
the test data set.
SVR is the SVM tool applied in regression analysis, and it estimates a continuous-valued
multivariate function (Huang, 2007). SVM possesses a firm mathematical basis on statistical
learning theory and the key objective of the Vapnik–Chervonenkis (VC) theory is the
characterisation of the generalised error instead of the error on specific data set (Verma et al.,
2016). This aids SVM to better generalise for unseen data. SVR tries to curtail the upper
bound on the generalisation error based on the Structural Risk Minimisation (SRM) principle
rather than Training Error Minimisation (TEM) (Awad and Khanna, 2015; Listiani, 2009).
SVR is generated from Support Vector Machine (SVM) through generalisation by
introducing an «-insensitive region around the function, termed the «-tube. It is formulated as
an optimisation problem with initially defining a convex «-insensitive loss function to be
minimised and recognising the flattest tube that includes the best of the training instances
(Awad and Khanna, 2015). Convex optimisation has a unique local minimisation solution.
The solution is achieved using appropriate numerical optimisation algorithms. A hyperplane
is represented by support vectors in SVR which are training samples that lie outside the Machine
boundary of the tube. learning and
SVM has three main attributes namely better generalisation capability, optimal global
solution using optimisation theory, and Kernel functions for nonlinearity (Huang, 2007).
construction
Hyperparameters contribute immensely to the performance of SVM. It is therefore project profit
substantial to set good value for them. The SVM parameter C and ε together with kernel
parameters k is known as hyperparameters, and optimise the performance generalisation.
Various kernel functions tolerate adjustment concerning training data. The training data 1501
influence the optimal hyperparameter setting and usually set manually since there is no
theory to delineate a good value (Huang, 2007).
The learning curve is required to appreciate the sensitivity of SVR relative to the size of the
training data set. The finding of fitting values for the hyperparameters through several
rounds of model development is a significant section in SVR. It requires more time when there
is more data to crunch to develop and validate a model. This is because if the repetition
required to discover a grid with a set of m cost values and n ε-values is m 3 n times, with an
average computational effort of t, then estimated duration of m 3 n 3 t will be required
(Listiani, 2009). The size of processing data dictates the definite duration taken for each
round. Hence, the computational time needs minimisation. The partial data set is enough to
produce a model due to the dependency of SVR on support vectors. The sensitivity of SVR is
performed through running numerous times with swelling sample numbers and observing
the learning effect gained from additional data by measuring Root-Mean-Squared Error
(RMSE) (Listiani, 2009).

Methodology
The methodology is presented and discussed under five key sub-sections to include data
collection and processing, feature selection, support vector regression, model evaluation and
model development and performance.

Data collection and processing

Historical data was used in the model development and validation. A total of 150 completed
projects which were financed with internally generated or donor funds within the past five
years (i.e. 2014–2018) formed the source of data for the study. The choice of the period was
settled on due to the price stability within this period. Eighty percent (80%) of these projects
were used in the development of the model while the remaining 20% were used in the
validation. The study explored the various financing options available to contractors in
project financing. This was motivated by the cost of capital on the various sources which
erodes profit. Therefore various combinations of these options were investigated to minimise
cost in the bid to maximise profit. A financing option consisting of contractor financing and
loan was employed as this is the commonest financial combination for construction projects
in Ghana. The variables used in this study were adopted from the findings of Adjei et al.
(2018) which established the most significant factors influencing profit on construction
projects in Ghana. Additionally, the cost contributions of the identified significant cash-flow
factors established by Adjei et al. (2019a) were also applied to the completed projects to
generate the dataset for the model development. This approach meant that no outliers were
recorded or observed in the dataset. The dataset was subsequently organised to suit the SVR
learning scheme in a spreadsheet.

Feature selection
Feature selection, also known as the variable selection, is selecting attributes in data that are
utmost appropriate to the predictive modelling problem under study. It improves accuracy
ECAM and efficiency and aims at reducing the dimensionality and noise in data sets (Ghojogh et al.,
28,5 2019; Roffo, 2018; Guajardo et al., 2007). Feature selection facilitates visualisation of data,
improves understanding of data (Guajardo et al., 2007) and minimises the risk of overfitting
(Hira and Gillies, 2015). Roffo (2018) stated that automatic feature selection, that is
computational variable selection techniques such as forward selection, backward elimination
and stepwise regression can be employed to remove inappropriate, redundant and noisy
information from data to facilitate better performance in learning. However, the human
1502 operator sometimes defines the potentially useful features in many learning domains (Roffo,
2018). In line with this, the features employed in this model were selected from the work of
Adjei et al. (2018) which used Principal Component Analysis (PCA) to establish the significant
cash-flow factors impacting on profit. These factors (features) were wages of labour and staff,
progress payment duration, bank interest rate and defective works. However, some of the
projects utilised in this study had some differences in the initial and final contract sums,
which established some variations; hence, variation was added to the selected features.
According to Windapo (2017), suppliers discounts or credit facilities are offered to enhance
the profit and cash flow of companies. In view of trying to predict profit from significant cash
factors, the study subsequently included a credit to enhance the prediction of profit.

Support vector regression (SVR)

SVR goes through two phases namely: learning and testing phases. The generated dataset of
150 projects was employed in these phases. According to Listiani (2009) and Verma et al.
(2016), a bigger proportion of data set is required in the learning phase than the testing phase;
hence, one could use anything around 70–30% as a rule of thumb. The study, therefore,
adopted an 80–20% data partition. Subsequently, data from the 150 projects obtained for the
study was divided into 120 (80% for hyperparameter selection phase and final training
phase) and 30 (20% for testing phase). The learning phase also goes through two steps
namely: the hyperparameter selection and training. Three different kernels (linear,
polynomial and radial basis function) were evaluated with the predictors of the generated
dataset to establish a suitable kernel for the development. The linear kernel was established
to be suitable as it resulted in the least RMSE among the others. At this stage, the suitable
hyperparameters were also generated automatically and subsequently employed in the
training of the dataset. The training is a process employed to enhance the prediction
capabilities of the model. It can be regarded as trying to find the set of parameters for a
specific function that can be used to make predictions given a set of independent variables.
The accuracy of any model prediction most of the time depends on the training it goes
through as well as the size of the training sample. In this study, a five-fold data partition was
used to train the model. A lot of variables may satisfy the constraints hence, the training
produced optimal model parameters. These are the minimum values for the respective
parameters and were obtained after several iterations. Figures 2–4 present the
hyperparameters selection and SVR learning scheme, the flowchart of model cross-
validation and the steps in predicting values, respectively.

Model evaluation
This is performed with the validation data and at this stage, the performance of the model is
critically examined. The performance of a model may not be satisfied after the initial training.
This demands the altering of some parameters and the training and evaluation processes
repeated until a good result is obtained. With the use of MATLAB for SVR, the parameters
available for training were the epsilon values, the kernel scale, the box constraint and
standardisation. The accuracy of the model, however, was always higher without
standardisation. A margin of error was therefore allowed to avoid overfitting as an
 Machine
learning and
construction
project profit

1503

Figure 2.
Hyperparameters
selection and SVR
learning scheme

overfitting model strives hard to attain zero error. This makes the model too sensitive since it
also fits outliers to produce a high variance and to prevent the model from missing relevant
information about the relationship between the data. Therefore, an effort was made to attain a
perfect model that was not too influenced and not too sensitive.

Model development and performance

The linear kernel was subsequently used in developing the model. According to Deleen and
Bonsu (2002), contractors mostly employ suppliers’ credit as an external source of financing
equipment, and this can be attributed to the initial high cost associated with equipment.
Windapo (2017) also stated that suppliers credit facilities improve company profits and cash
flow. It can, therefore, be inferred that projects cannot be executed successfully without
credit, irrespective of duration, because of payment irregularities that impact cash flow. This
necessitated the inclusion of credit to loan and contractor finance to develop the model. Hence,
the finance options settled for the model developed in the study were contractor finance, loan
and credit.
The Mean Absolute Percentage Error (MAPE) was employed to evaluate the
performance of the developed model. Results obtained from the MAPE expression is
interpreted using Table 2. Liu and Pan (2012) reported that MAPE is the most effective
estimation index and formulates related evaluation standards. The sensitivity of a model is
an important contemplation in appraising the performance of that model. The sensitivity
analysis of a study deals with how the uncertainty in the output of a mathematical model
can be divided and assigned to different sources of uncertainty in its inputs. Since the
projects used in the model development and validation were donor and internally generated
funded projects, it is presumed that the funds for the projects were readily available.
Therefore, a delay is assumed to be minimised. It has been reported that the parameters of
SVR, that is epsilon (e), cost or error penalty (C) and the kernel type influence prediction
performance (Priyadarshini et al., 2011; Ustuner et al., 2015). Additionally, the sensitivity
index can be used to determine the degree to which the independent variables impact the
dependent variable (Hamby, 1994). In achieving this, the prediction of a specific variable is
determined at its maximum and minimum values, keeping the others constant. This index
has a value between 0 and ± 1. The sensitivity of a variable is determined with the closeness
of the index to 1 (Hamby, 1994). The study, therefore, adopted the sensitivity index to
ECAM Start
28,5

Break data into K-folds

1504 Here K = 5
Set n = 1

Train with all but nth fold

Test with nth fold

Yes
Is n = 1?

No

Yes Set n = n + 1
Set ε = εn Is ε > εn ?

No

Select nth model Is n = k?

No

Yes
Figure 3.
Flowchart of model
cross-validation End

determine the degree to which the independent variables (cash-flow factors) impacted the
dependent variable (profit).

Results
Completed institutional building projects with internally generated and donor funding were
used for this study. The established cost contributions of the identified significant
quantifiable cash-flow factors were subjected to 150 completed projects under study to
determine the respective cost allocation to the projects. These consequently constituted the
dataset for the model development and validation. It was identified that some of the projects
had variations; hence, variation was included in the variables for model development and
further predictions. Funds for the projects under study were presumed to be readily available;
therefore, payment schedules were not unduly delayed. It can, therefore, be inferred that
delay in the projects was minimised which subsequently had less influence on the variation
and expected profit of the projects obtained.
Machine Learning Toolbox in MATLAB was used to develop the model, and this includes
functions to perform various analyses on multivariate data. Support vector regression
 Machine
learning and
construction
project profit

1505

Figure 4.
Steps in predicting
values

MAPE Interpretation

<10 Highly accurate forecasting

10–20 Good forecasting
20–50 Reasonable forecasting Table 2.
>50 Inaccurate forecasting Interpretation of
Source(s): Liu and Pan (2012), Jos
e et al. (2013) MAPE values

analysis uses an optimal line, hyperplane drawn across the training data. This line is usually
straight but might appear to be distorted when the number of dimensions constituting the
hyperplane goes beyond three. The performances of the various kernels were assessed, and
the linear kernel was settled on to be the best performing kernel for the model development as
it generated the least RMSE.
The linear kernel was subsequently used in developing the model. In subjecting the
financing option datasets to develop the model, the coefficients of the variables in the model
shown in Table 3 were attained.
The coefficients generated and the independent variables are represented in 1 3 n and
n 3 1 vectors or matrixes, respectively. The prediction which is a scalar (1 3 1) is the
product of 1 3 n and n 3 1. Therefore, the general expression for the linear kernel which is
the inner product plus a constant was thereafter performed to generate the model shown in
equation (2).
1
Profit ¼ ðβ1ICS þ β2Vr þ β3Lb þ β4Icf þ β5Iln þ β6Icr þ β7LD þ β8DW þ bÞ (2)
k
In assessing the measure of goodness of fit, a resulting value of 1 was obtained as the
determinant (R2) of the established model with a 1% margin of error. This indicates the
closeness of the data to the hyperplane. This further explains that the model predictions
perfectly fit the data and explains all the variability of the response data around its mean.
Average predictive accuracy of 73.66% was achieved with all the variables on the different
ECAM projects in validation. This result gives a MAPE value of less than 30% (26.18%). In reference
28,5 to Table 2 as indicated by Liu and Pan (2012) and Jos
e et al. (2013), reasonable forecasting is
said to be achieved when MAPE value range between 20 and 50%. Therefore, it can be
inferred that; a reasonable prediction of profit is attained with this model.

The sensitivity of the model

1506 From Table 4, it was identified that reasonable predictions were obtained with all the
maximum and minimum values except for the minimum for labour which produces a good
prediction when the sensitivity index was employed. Furthermore, the prediction results
exhibited an upward trend of prediction from the maximum variable to the minimum.
However, it was the reverse case for a loan. Labour and contractor finance generated SI
values of 0.320 and 0.03, respectively as shown in Table 4. Upon introducing crediting,
delay, and defective works separately, 0.002, 0.012, and 0.008 respectively, were the SI
results obtained. Contrary to the trend of results, loans generated a positive SI value of 0.04. It
can therefore be deduced that maximising this input variable whiles holding others constant
demonstrated an enhancement in predicting profit. However, this is the reverse in the case of
a loan. Although loans are associated with high interest paid on them, its acquisition
improves the profit earnings.

Discussion
The study has been able to establish a novel profit prediction model using SVR with
significant cash-flow factors as predictors. These predictors account for 100% of the variance
in the profit generation as a result of a perfect correlation obtained from R2 5 1. The approach
adopted to generate the dataset for developing the model meant that there were no data
outliers. Data outliers have a significant impact on the accuracy of the predictive model.
However, the proposed model resulted in MAPE less than 30%, indicating a reasonable
prediction. In Park and Kim’s (2011) Bayesian profit predicting models of construction
projects, technical factors were considered as predictors and these also accounted for at least

Variables Symbol Coefficients

Initial contract sum (Ics) β1 24498.22

Variation (Vr) β2 24798.70
Labour (Lb) β3 13186.93
Interest on contractor finance (Icf) β4 1300.02
Interest on loan (Iln) β5 2351.68
Table 3. Interest on credit (Icr) β6 375.22
Determined Loss due delay (Ld) β7 1658.56
coefficients of Defective works (Dw) β8 1740.18
independent variables Bias (B) B 0.00207
of model Kernel scale (K) K 172803.4

Variable Labour Loan Contractor finance Credit Delay Defective works

Table 4. Dmax 60.67 75.77 71.55 73.66 73.05 73.41

Sensitivity index of Dmin 80.65 73.10 74.06 73.84 74.02 73.97
respective variables SI 5 Dmax-Dmin Dmax 0.329 0.04 0.03 0.002 0.013 0.008
76.9% of the variance in the dependent variable. However, this study failed to evaluate the Machine
sensitivity of the predictors. Additionally, El-Kholy (2014) developed a multi-objective fuzzy learning and
linear programming (FLP) model for cash-flow management that reduced the optimum value
of final cash balance by a percentage between zero and 19% in FLP from the corresponding
construction
ideal value in the crisp linear programming (LP) model. This resulted in an optimum value of project profit
initial cash balance that increased between 6.4 and 15.6% in FLP from the corresponding
ideal value in the LP. This study also failed to test for the sensitivity of the predictors. The
multi-period dynamic model by Jiang (2012) permits operators to outline the cash-flow 1507
planning horizon, as well as forecast and maximise the cash balance for the planning horizon.
Subsequently, the cash-flow model was established to be sensitive to a retained percentage,
progress payment schemes. Huang (2007) further affirms the predictability and performance
of SVR in the determination of the cost of complex products at the early phases of designing.
With reference to previous studies, the Choi et al. (2013) profit model considered labour
productivity per construction worker whereas Jiang et al.’s (2011) model concentrated on
financial market constraints such as long- and short-term loans from banks, gains on excess
cash deposit and minimum cash reserves of a project. El-Kholy (2014) further modified Jiang
et al.’s (2011) model and incorporated delay payment for only one period. The developed
model in this study incorporated variables from El-Kholy’s study with the inclusion of
variation, initial contract sum and defective works.
Choi et al. (2013) employed predicted error sum of square (PRESS) and the sum of squared
error (SSE) to evaluate the accuracy of the developed exponential profit model which
considered labour productivity per construction worker. The study stated that, the closeness of
PRESS to SSE indicates the model possesses significant predictability. However, if the
difference is several times larger, then the model has an issue with validation. The values of
PRESS and SSE obtained in the study were 3.265 and 2.819 respectively indicating a significant
prediction. The accuracy of Park and Kim’s (2011) model was evaluated with the differences of
BRA model against the actual margin of profit ranged from 1.82 to 0.01%, while that of the
multiple linear regression (MLR) ranged from 0.75 to 4.71%. The compared results clearly
reflected the superiority of BRA over MLR. Bee-Hua’s (2010) study also examined the
combination of neural networks (NNs) and genetic algorithms (GAs) to forecast residential
construction demand which resulted in reduction of average MAPE from about 6% to a mere
1%. However, the NNs and Gas models generated accurate forecasts, since their respective
MAPE values consistently fell within the acceptable limit of 10%. This established an enhanced
performance produced from the use of a hybrid technique. Additionally, Emsley et al.’s (2010)
neural network model to predict total construction costs demonstrated the capability of the
neural network to model the nonlinearity in the data. This was revealed in an average MAPE of
16.6% comparable to the traditional estimate which reported between 20.8 and 27.9%. More so,
Hua and Pin’s (2010) Boxjenkins model to predict the construction demand, price and
productivity employed RMSE and MAPE to evaluate the model accuracy. All the developed
models generated RMSE which were consistently smaller than the standard errors and MAPE
also fell within the acceptable limit of 10% with demand model possessing the least.
The model developed in this study has the independent variables accounted for 100%
variance explained in the dependent variable. It can be established that the SVR performs
better in prediction accuracy than other techniques except for hybrid techniques. Although
much work has not been conducted on SVR concerning productivity and profit, it can be
inferred from these discussions that this study affirms the superiority of SVR in prediction
and accuracy in predicting profit to enhance the productivity of the construction industry.
The utilisation of loans contributes to the early completion of projects for certification and
onward payment. Therefore, this reduces capital lock-up should the project completion be
delayed when funds are not available to finance for regular and prompt payments. It can be
established that labour is sensitive to the prediction of profit compared to the other independent
ECAM variables. Apart from the loan, it can be deduced that any upsurge of the variable will
28,5 negatively impact on profit prediction. Sensitivity was also performed on variation since it is a
variable employed to predict profit. This variable was employed in the development of the
model because some of the projects used in the development and validation of the model had
recorded significant variation. Since the coefficient of variation is positive, any addition or
omission to a contract will impact on the profit positively or negatively respectively. About the
sensitivity index results obtained, projects should be implemented with minimum labour, credit
1508 and contractor finance to enhance profit. However, financing with a loan can be encouraged to
facilitate the early completion of projects to receive payment promptly to improve profit
generation whenever the funds for payment are readily available.

Conclusion and recommendation

Construction productivity rate is the origin of the accurate estimating of the cost and time
necessary to complete a project. Since enhanced productivity helps contractors and project
owners to realise their profitability levels, productivity has been considered an important
predictor of profitability. The threats posed by productivity to the profitability of
construction organisations have called for several solutions from industry researchers
across the globe. However, aside from productivity, cash flow has also been identified as a
potential threat, with effective cash flow management considered as key for construction
organisations to survive in the already competitive construction industry.
This study was conducted to develop and test the sensitivity of a Machine Learning
Support Vector Regression Algorithm (SVRA) to predict construction project profit in Ghana.
Historical data was used in the model development and validation. This data was obtained
from 150 institutional projects that were executed within the past five years (i.e. 2014–2018).
Eighty percent (80%) of the data from the 150 projects was used at hyperparameter selection
and final training phases of the model development and the remaining 20% for testing the
model. Data were analysed using MATLAB for Support Vector Regression. The findings
from the study suggest that the predictions from the developed model perfectly fitted the
data, thereby explaining all the variability of the response data around its mean. The findings
further suggest that on average predictive accuracy of 73.66% was achieved with all the
variables on the different projects in validation, with the developed SVR model being
sensitive to labour and loan.
This study has led to the development of a Support Vector Regression Algorithm that will
serve to prompt or warn contractors before the commencement and during the execution of
projects about profit performance. There are significant theoretical and practical implications
from the findings of this study. Theoretically, the study contributes to the ever-growing
literature on the application of machine learning in predicting construction project profit.
Literature has reported numerous models and techniques that have been developed and used
in different contexts. The SVRA model developed in this study contributes its quota by
potentially filling the profit measurement gap with the incorporation of variation, defective
works, and labour to financial constraints, which have been the variables used in previous
studies. Practically, the model should aid contractors to predict profit on completion before
the commencement of projects. This will equip contractors seeking to enhance profit by
helping them make informed and objective decisions towards maximisation of profit. It will
provide contractors with relevant signals on the influence of any change in the significant
variables used in the determination of profit because of the client (variations or payment
issues), or the contractor (labour management).
Despite the useful outcome of the study, it was limited to only contractors who are well
resourced and execute large volumes of works. It also only looked at quantifiable cash-flow
factors and further considered a year duration of interest payment on the various sources of
finance. It will be useful to conduct studies to evaluate the accuracy of the model over the long
term when payment is unduly delayed with the incorporation of interest paid on delayed Machine
payment. Again, similar studies on projects undertaken by smaller contractors who are not learning and
well resourced could unearth other significant cash-flow factors from their perspective,
affecting profit predictions by small contractors who are in the majority, especially when one
construction
considers construction industries in developing countries. project profit

References 1509
Achenef, T.K. (2016), “Factors affecting profitability: the case of Ethiopian selected private
construction companies”, A thesis submitted to, St. Mary’s University School of Graduate
Studies, Addis Ababa.
Adjei, E.A., Fugar, F.D.K., Adinyira, E., Edwards, D.J. and P€arn, E.A. (2018), “Exploring the significant
cash flow factors influencing building projects profitability in Ghana”, International Journal of
Construction Engineering and Management, Vol. 7 No. 1, pp. 35-46.
Adjei, E.A., Fugar, F.D.K. and Adinyira, E. (2019a), “Establishing the cost contribution of significant
cash flow factors impacting on building projects’ profitability”, in Marshall-Ponting, A., Ji, Y.,
Keraminiyage, K.P., Lee, A., Poppelreuter, T., Swan, W., Udeaja, C.E. and van Dijk, H. (Eds),
2019, Conference Proceedings of the 14th International Postgraduate Research Conference 2019:
Contemporary and Future Directions in the Built Environment, ISBN 9781912337309, University
of Salford, Salford, pp. 25-37.
Adjei, E.A., Fugar, F.D.K. and Adinyira, E. (2019b), “A review of appropriate tool to predict
profitability of building projects using established significant cash flow factors”, in Marshall-
Ponting, A., Ji, Y., Keraminiyage, K.P., Lee, A., Poppelreuter, T., Swan, W., Udeaja, C.E. and van
Dijk, H. (Eds), 2019, Conference Proceedings of the 14th International Postgraduate Research
Conference 2019: Contemporary and Future Directions in the Built Environment, ISBN
9781912337309, University of Salford, Salford, pp. 38-48.
Akintoye, A. and Skitmore, M. (1991), “Profitability of UK construction contractors”, Construction
Management and Economics, Vol. 9 No. 4, pp. 311-325.
Arditi, D., Koksal, A. and Kale, S. (2000), “Business failures in the construction industry”, Engineering,
Construction and Architectural Management, Vol. 7, pp. 120-132.
Asante, J.A. (2014), “Financial distress related causes of project delays in the Ghanaian construction
industry”, Unpublished, available at: [Link]
ASANTE%20JOYCELINE%[Link] (Accessed July 2019).
Avetisyan, E.H., Tesha, D.N.G.A.K., Lello, D.S. and Mtitu, F.S. (2020), “Profit maximization strategies
employed by the small and medium size building contractors in Dar-Es-Salaam, Tanzania”,
International Journal of Engineering and Management Research, Vol. 10 No. 1, doi: 10.31033/
ijemr.10.1.17.
Awad, M. and Khanna, R. (2015), “Support vector regression”, in Efficient Learning Machines, Apress,
Berkeley, CA, pp. 67-80.
Aziz, R.F. (2013), “Optimising strategy for repetitive construction projects within multi-mode,
resources”, Alexandria Engineering Journal, Vol. 52, pp. 67-81, doi: 10.1016/[Link].2012.11.003.
Barbosa, P.S.F. and Pimentel, P.R. (2010), “A linear programming model for cash flow management in
the Brazilian Construction Industry”, Construction Management and Economics, Vol. 19 No. 5,
pp. 469-479, doi: 10.1080/01446193.2001.9709623.
Bee-Hua, G. (2010), “Evaluating the performance of combining neural networks and genetic
algorithms to forecast construction demand: the case of the Singapore residential sector”,
Construction Management and Economics, Volume, Vol. 18 No. 2, pp. 209-217, doi: 10.1080/
014461900370834.
nski, W. (2012), “The influence of liquidity on profitability of polish construction
Bolek, M. and Wili

sector companies”, Financial Internet Quarterly e-Finance, Vol. 8 No. 1, pp. 38-52, available at:
[Link]
ECAM Bowman, C. (2014), “Where does profit come from?”, available at: [Link]
or [Link] (accessed 21 August 2014).
28,5
Buertey, J.I.T. and Adjei-Kumi, T. (2012), “Cash flow forecasting in the construction industry: the case
of Ghana”, Pentvars Business Journal, Vol. 6 No. 1, pp. 65-81, available at: [Link]
com/articles//january-2012/[Link].
Chendroyaperumal, C. (2009), “Profit theories: modern and Indian”, (Working Paper), available at:
[Link] (accessed December 2020).
1510
Cheng, M.-Y., Hoang, N.-D. and Wu, Y.-W. (2015), “Cashflow prediction for construction project using
a novel adaptive time-dependent least squares support vector machine inference model”,
Journal of Civil Engineering and Management, Vol. 21 No. 6, pp. 679-688.
Choi, K., Haqui, M., Lee, H., Cho, Y. and Kwak, Y. (2013), “Macroeconomic labour productivity and its
impact on firm’s profitability”, Palgrave Macmillan, Journal of the Operational Research Society,
Vol. 64, pp. 1258-1268, doi: 10.1057/jors.2012.157.
Deelen, L. and Bonsu, K.O. (2002), Equipment Finance for Small Contractors in Public Work
Programmes, International Labour Organization, Geneva.
El-Kholy, A.M. (2014), “A multi-objective fuzzy linear programming model for cash flow managemen”,
International Journal of Engineering Research and Applications, Vol. 4 No. 8, pp. 152-163,
(Version 3), available at: [Link]
pdf (accessed July 2018).
Emsley, M.W., Lowe, D.J., Duff, A.R., Anthony, H. and Hickson, A. (2010), “Data modelling and the
application of a neural network approach to the prediction of total construction costs”,
Construction Management and Economics, Vol. 20 No. 6, pp. 465-472, doi: 10.1080/
01446190210151050.
Enshassi, A., Mohamed, S. and Abushaban, S. (2009), “Factors affecting the performance of
construction projects in the Gaza Strip”, Journal of Civil Engineering and Management, Vol. 15
No. 3, pp. 269-280, doi: 10.3846/1392-3730.2009.15.269-280.
Ghojogh, B., Samad, M.N. and Mashhadi, S.A. (2019), “Feature selection and feature extraction in
pattern analysis: a literature review”, available at: [Link]
332960460_Feature_Selection_and_Feature_Extraction_in_Pattern_Analysis_A_Literature_
Review (accessed November 2020).
Guajardo, J., Weber, R. and Miranda, J. (2006), “A forecasting methodology using support vector
regression and dynamic feature selection”, Journal of Information and Knowledge Management,
Vol. 5 No. 4, pp. 329-335, doi: 10.1142/S021964920600158.
Gundecha, M.M. (2013), “Study of factors affecting labour productivity at building construction
project in the USA”, Web Survey, available at: [Link]
repository/bitstream/handle/10365/22772/Gundecha_Mahesh.pdf?sequence51 (accessed
July 2017).
Halim, M.S.B.A. (2014), “Unrealistic profit experienced by Bumiputera entrepreneurs in the
Malaysian construction industry”, Advances in Environmental Biology, Vol. 8 No. 9,
pp. 489-496, 1995-0756.
Hamby, D.M. (1994), “A review of techniques for parameter sensitivity analysis of environmental
models”, Environmental Monitoring and Assessment, Vol. 32, pp. 135-154, available at: http://
[Link]/viewdoc/download?doi510.1.1.224.7813&rep5rep1&type5pdf.
Harris, F. and McCaffer, R. (2005), Modern Construction Management, Blackwell Publishing,
New Jersey, p. 118138.
Haseeb, M., Bibi, A. and Rabbani, W. (2011), “Problems of projects and effects of delays in the
construction industry of Pakistan”, Australian Journal of Business and Management Research,
Vol. 1, pp. 41-50.
Hillebrandt, P.M. (2000), Economic Theory and the Construction Industry, Palgrave Macmillan,
New York.
Hillebrandt, P.M., Cannon, J. and Lansley, S. (1995), The Construction Company in and Out of Machine
Recession, Macmillan, London.
learning and
Hira, Z.M. and Gillies, D.F. (2015), “A review of feature selection and feature extraction methods
applied”, Advances in Bioinformatics, Vol. 2015, pp. 1-13, 198363, doi: 10.1155/2015/198363.
construction
Hofstrand, D. (2013), “Cash flow and profitability are not the same”, Ag Decision Maker, File C5-213,
project profit
available at: [Link]/agdm (accessed 15 January 2021).
Hua, G.B. and Pin, T.H. (2010), “Forecasting construction industry demand, price and productivity in 1511
Singapore: the Boxjenkins approach”, Construction Management and Economics, Vol. 18 No. 5,
pp. 607-618, doi: 10.1080/014461900407419.
Huang, G. (2007), “Cost modelling based on support vector regression for complex products during the
early design phases”, Unpublished.
Hughes, W., Hillebrandt, P. and Murdoch, J. (2000), “The impact of contract duration on the cost of
cash retention”, Construction Management and Economics, Vol. 18, pp. 11-14, doi: 10.1080/
014461900370906.
Jaafar, M. and Abdul-Aziz, A.R. (2005), “Resource-based view and critical success factors: a study on
small and medium sized contracting enterprises (SMCEs) in Malaysia”, International Journal of
Construction Management, Vol. 5 No. 2, pp. 61-77.
Jiang, A. (2012), “Negotiating construction contracts through practical cash flow planning
and analysis model”, International Journal of Construction Management, Vol. 12 No. 2,
pp. 23-33, doi: 10.1080/15623599.2012.10773188.
Jiang, A., Issa, R.R.A. and Malek, M. (2011), “Construction project cash flow planning using the Pareto
optimality efficiency network model”, Journal of Civil Engineering and Management, Vol. 17
No. 4, pp. 510-519, doi: 10.3846/13923730.2011.604537.
Jos
e, J.M.M., Alfonso, P.P., Albert, S.A. and Berta, C.B. (2013), “Using the R-MAPE index as a resistant
measure of forecast accuracy”, Psicothema, Vol. 25 No. 4, pp. 500-506, doi: 10.7334/
psicothema2013.23.
Kaka, A.P. and Cheetham, D.W. (1997), “The effect of some tendering and payment strategies on
contractors’ financial performance”, in 13th Annual ARCOM Conference, pp. 15-17.
Khosrowshahi, F. and Kaka, A. (2007), “A decision support model for construction cash flow
management”, Computer-Aided Civil and Infrastructure Engineering, Vol. 22, pp. 527-539,
available at: [Link]
referer5wolandtracking_action5preview_clickandshow_checkout51andpurchase_
[Link].comandpurchase_site_license5LICENSE_DENIED.
Kolisetty, V. and Rajput, D.S. (2020), “A review on the significance of machine learning for data
analysis in big data”, Jordanian Journal of Computers and Information Technology (JJCIT),
Vol. 06 No. 1, doi: 10.5455/jjcit.71-1564729835.
Langford, D.A. and Chan, K.K. (1987), “Labour only subcontracting – the changing position”,
Construction Management and Economics, Vol. 5 No. 1, pp. 31-33.
Lee, F.B. (2009), “Factors affecting the profitability of construction companies in Hong Kong”,
Unpublished, doi: 10.5353/th_b4727769 (accessed November 2017).
Listiani, M. (2009), “Support vector regression analysis for price prediction in a car leasing
application”, Unpublished, available at: [Link]
pw-and-m/source/papers/2009/[Link] (accessed November 2017).
Liu, S. and Pan, N. (2012), “’Construction firms’ operation management performance measurement
model”, Journal of Statistics and Management System, Vol. 15 No. 6, pp. 627-642, doi: 10.1080/
09720510.2012.10701645.
Liu, Y., Zayed, T. and Li, S. (2009), “Cash flow analysis of construction projects”, 2nd International/
8th Construction Specialty Conference, pp. 1-8, available at: [Link]
72b1/[Link] (accessed July 2017).
ECAM Long, R.J. (2015), Contract Scheduling Provisions, Long International, Littleton, CO.
28,5 Mahamid, I. (2011), “Causes of contractors’ failure: contractors’ view”, 2nd International Conference on
Construction and Project Management IPEDR, Vol. 15. ISBN: 0047735945.
Majer, R., Ellingerov
a, H. and Gasparik, J. (2020), “Methods for the calculation of the lost profit in
construction contracts”, Buildings, Vol. 10, pp. 1-13.
Malik, H. (2011), “Determinants of insurance companies profitability: an analysis of insurance sector
1512 of Pakistan”, Academic Research International, Vol. 1 No. 3, pp. 315-321.
Marchal, J. (1951), “The construction of a new theory of profit”, The American Economic Review,
Vol. 41 No. 4, pp. 549-565.
McGuigan, J.R., Moyer, R.C. and Harris, F.H. (2017), Managerial Economics: Applications, Strategy, and
Tactics, 14th ed., South-Western, Cengage Learning, London, ISBN-13: 978-1-4390-7923-2.
Mohsin, M.A., Alnuaimi, A. and Tobi, S.A. (2014), “Contractual implications of cash flow on owner and
contractor in villa construction projects”, International Journal of Research in Engineering and
Technology, Vol. 3 No. 4, pp. 442-447, doi: 10.15623/ijret.2014.0304079.
Moon, S., Ham, N., Kim, S., Hou, L., Kim, J. and Kim, J. (2020), “Fourth industrialisation-oriented offsite
construction: case study of an application to an irregular commercial building”, Engineering
Construction and Architectural Management, Vol. 27 No. 9, pp. 2271-2286, doi: 10.1108/ECAM-
07-2018-0312.
Odeyinka, H.A., Lowe, J. and Kaka, A.P. (2013), “Artificial neural network cost flow risk assessment
model”, Construction Management and Economics, Vol. 31 No. 5, pp. 423-439, doi: 10.1080/
01446193.2013.802363.
Olatunji, A.A. (2010), “Influences on construction project delivery time”, A Thesis Submitted at, The
Nelson Mandela Metropolitan University, Port Elizabeth.
Omopariola, E.D., Windapo, A., Edwards, D.J. and Thwala, W.D. (2020), “Contractors’ perceptions of
the effects of cash flow on contruction projects”, Journal of Engineering, Design and Technology,
Vol. 18 No. 2, pp. 308-325.
Owalabi, J.D., Amusan, L.M., Oloke, C.O., Olusanya, O. and Tunji-Olayeni, P. (2014), “Causes and
effects of delay on project construction delivery time”, International Journal of Education and
Research, Vol. 2 No. 4, pp. 197-208, available at: [Link]
(accessed July 2017).
Oyewobi, L.O., Windapo, A.O. and Cattell, K.S. (2014), Competitiveness of Construction Organisations
in South Africa, Construction Research Congress, ASCE, Atlanta, pp. 2063-2073.
Pansegrauw, M. (2019), “Revenue is vanity, profit is sanity but cash is king”, available at: https://
[Link]/blog/podcast/revenue-is-vanity-profit-is-sanity-but-cash-is-king/
(accessed 21 January 2021).
Park, S. and Kim, S. (2011), “A strategy Bayesian Model to predict profit of construction projects”,
Architectural Research, Vol. 13 No. 3, pp. 49-56.
Park, H.K., Han, S.H. and Russell, J.S. (2005), “Cash flow forecasting model for general contractors
using moving weights of cost categories”, Journal of Management in Engineering, Vol. 21 No. 4,
pp. 164-172, doi: 10.1061/(ASCE)0742-597X(2005)21:4(164).
Peterson, S.J. (2009), Construction Accounting and Financial Management, 2nd ed., Pearson, London.
Pham, D.T. and Afify, A.A. (2005), “Machine-learning techniques and their applications in
manufacturing”, Journal of Engineering Manufacture, Vol. 219 No. 5, pp. 395-412, doi: 10.
1243/095440505X32274.
Priyadarshini, D., Acharya, M. and Mishra, A.P. (2011), “Link load prediction using support vector
regression and optimisation”, International Journal of Computer Applications, Vol. 24 No. 7, pp. 22-25.
Ramachandra, T. and James Rotimi, J.O.B. (2011), “The nature of payment problems in the New
Zealand construction industry”, Australasian Journal of Construction Economics and Building,
Vol. 11 No. 2, pp. 22-33, doi: 10.5130/AJCEB.v11i2.2171.
Reich, R. (2000), The Wealth of Nations, The Modern Library United State of America, New York. Machine
RICS (2011), “Cash flow forecasting”, RICS Practice Standards, UK 1st ed., Guidance note, ISBN 978 1 learning and
84219 687 8.
construction
Roffo, G. (2018), “Feature selection library (MATLAB Toolbox)”, available at: [Link] project profit
[Link].
Sambasivan, M. and Soon, Y.W. (2007), “Causes and effects of delays in Malaysian construction
industry”, International Journal of Project Management, Vol. 25, pp. 517-526, doi: 10.1016/j. 1513
ijproman.2006.11.007.
Sanyal, R. (2019), “Profit theory 1”, Book Chapter In Book: Planned Economy and growth Publisher:
S Chand, available at: [Link]
Sapankevych, N. and Sankar, R. (2009), “Time series prediction using support vector machines: a
survey”, IEEE Computational Intelligence Magazine, Vol. 4 No. 2, pp. 24-38.
Seo, K., Soh, J. and Sharma, A. (2018), “Do financial constraints affect the sensitivity of investment to
cash flow? New evidence from franchised restaurant firms”, Tourism Economics, Vol. 24 No. 6,
pp. 645-661.
Sepp€al€a, C.R. (2005), Contractors Claims Under the FIDIC Contracts for Major Works, International
Construction Contracts and Dispute Resolution Co-Hosted by ICC and FIDIC In partnership
with The Cairo Regional Centre for International Commercial Arbitration, Unpublished.
Snyman, T. and Smallwood, J. (2017), “Improving productivity in the business of construction”,
Procedia Engineering, Vol. 182, pp. 651-657.
Spedding, A.H. (1977), “Trends and prices in primary school building”, PhD thesis, University of
Leeds, Leeds.
Su, Y. and Lucko, G. (2015), “Synthetic cash flow model with singularity functions for unbalanced
bidding scenarios”, Construction Management and Economics, Vol. 33 No. 1, p. 3554, doi: 10.
1080/01446193.2015.1012527.
Tracy, J.A. and Tracy, T. (2015), “The relationship between cashflow and profit in business”, available
at: [Link]
profit-in-business-2/ (accessed 15 January 2021).
Tunji, S.T., Adebayo, O.S. and Tolulope, O.O. (2015), “Impact of gearing on performance of
companies”, Arabian Journal of Business and Management Review, Vol. 3 No. 1, pp. 68-80.
Usman, N., Sani, A. and Tukur, I.R. (2016), “Exploring the use of financial capacity as apredictor of
construction company corporate performance: evidence from South Africa”, International
Journal of Research and Development Organisation, Vol. 2 No. 2, pp. 2455-2466.
Ustuner, M., Sanli, F.B. and Dixon, B. (2015), “Application of support vector machines for land use
classification using high-resolution RapidEye images: a sensitivity analysis”, European Journal
of Remote Sensing, Vol. 48, pp. 403-422, doi: 10.5721/EuJRS20154823.
Vapnik, V. (1998), “The support vector method of function estimation”, in Nonlinear Modeling,
Springer, Boston, MA, pp. 55-85.
Vercellone, C. (2008), The New Articulation of Wages, Rent and Profit in Cognitive Capitalism, The Art
of Rent, Queen Mary University School of Business and Management, London, Chalshs-
00265584D.
Verma, N., Das, S. and Srivastava, N. (2016), “Multiple kernel support vector regression with higher
norm in option pricing”, International Journal of Applied Engineering Research, Vol. 11, No. 5,
pp. 3168-3174.
Windapo, A. (2017), “Determinants of the sustainability and growth of construction contractors”,
pp. 1-13, doi: 10.20944/preprints201707.0091.v1.
Wright, M.G. (1977), “Profit and competition: profitability”, Building Technology and Management, s.l.,
s.n, pp. 4-6.
ECAM Yasamis, F., Arditi, D. and Mohammadi, J. (2002), “Assessing contractor quality performance”,
Construction Management and Economics, Vol. 20 No. 3, pp. 211-223, doi: 10.1080/
28,5 01446190110113693.
Yean, F., Ling, Y. and Liu, M. (2005), “Factors considered by successful and profitable contractors in
mark-upsize decision in Singapore”, Building and Environment, Vol. 40, pp. 557-1565, doi: 10.
1016/[Link].2004.12.001.
Zhu, F., Hu, H., Xu, F. and Tang, N. (2019), “Predicting profit performance of international
1514 construction projects”, Proceedings of the 2019 IEEE IEEM Conference, pp. 445-449.

Corresponding author
Emmanuel Adinyira can be contacted at: [Link]@[Link]

For instructions on how to order reprints of this article, please visit our website:
[Link]/licensing/[Link]
Or contact us for further details: permissions@[Link]

View publication stats