IConCEES-2023 IOP Publishing

IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

Predictive model for corrective maintenance costs:

Empowering decision-making in building renovation

A Hauashdh1’*, S Nagapan 1, J Jailani 1 and S Alzaeemi 2

1
Faculty of Civil Engineering and Built Environment, Universiti Tun Hussein Onn Malaysia,
Batu Pahat, Johor, 86400, Malaysia
2
Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, Batu
Pahat, Johor Malaysia, 86400, Malaysia
*Corresponding author: [email protected]

Abstract. The prediction of corrective maintenance costs is significant given the predominant
use of a corrective maintenance approach in building maintenance. Accurately estimating the
costs associated with corrective maintenance at an early stage holds substantial implications for
cost allocation, maintenance budgeting, cost effectiveness, and efficient planning, all of which
are vital factors contributing to the overall success of building maintenance. However, the
utilization of historical data to predict future maintenance costs remains underutilized. To
contribute to this gap, this study aims to develop a prediction model for the number of building
defects and their associated costs based on past data of defects and building age. The study
encompasses 40 buildings and employs regression analysis to develop a predictive model. The
predictive model was coded in Python to validate and ensure logical outputs and alignment with
expected outcomes while also utilizing the Pearson product-moment correlation coefficient
between variables and model output accuracy. The resulting model can provide logical
outcomes, enabling accurate predictions of corrective maintenance costs for each building.
Furthermore, it assists decision-making regarding cost considerations, such as determining
whether an aging building should be renovated or if repairing specific defects based on a
corrective approach is more beneficial. In summary, this study contributes to enhancing
maintenance planning and informed decision-making, providing significant benefits for
maintenance cost estimation, and building renovation decisions.

1. Introduction
The main purpose of building maintenance is to prevent or repair defects, and the maintenance costs
primarily arise from the initial efforts to prevent or repair defects in buildings. This cost consists of
direct costs, which encompass the cost of defects, and indirect costs associated with operating expenses
[1]. Therefore, accurately estimating these costs is crucial for effective maintenance planning. However,
there is a lack of maintenance cost prediction approaches [2], as predicting the cost of defects is complex
and challenging due to the need to forecast their occurrence within the building [1,3]. Further, it is
challenging due to the limited access to historical maintenance data, hindering the generation of reliable
forecasts for yearly costs [4].
Most building maintenance organizations allocate their building maintenance budget annually,
primarily relying on a corrective maintenance strategy [1,5]. Besides, with the increase in building age,
the cost of building defects becomes high, and the building becomes more susceptible to being put out
of service for many days each year [6,7]. This poses a significant challenge, particularly when older
buildings need to be continued to maintain by corrective approach or to be renovated. Thus, having

Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
IConCEES-2023 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

knowledge of the number of building defects and their associated maintenance costs allows for
determining whether a building should undergo renovation and identifying the optimal timing for the
renovation. Accordingly, this data assists in identifying buildings that may have significant maintenance
expenses due to the frequency with which faults arise each year. Additionally, the utilization of Building
Information Modeling (BIM) and other maintenance technology tools has gained traction in several
building maintenance organizations nowadays [2,8]. These tools enable the establishment of
comprehensive systems that record and archive previous maintenance works and related data for each
building. This includes storing information about past building defects and their associated details for
each building [6], which enables the utilization of this data to predict the cost of maintenance and plan
properly for maintenance. While past building defects data is a valuable resource [1,9,10], its use in
enhancing future maintenance plans is still primarily limited to transparency purposes, serving as proof
of completed work. However, further analysis is required to effectively utilize this data in enhancing
building maintenance planning. Consequently, this study employs past building defects data to improve
the forecasting of maintenance costs. The study presents a prediction model for estimating the costs of
corrective maintenance in buildings. This model can assist building maintenance organizations and other
stakeholders estimating the annual maintenance costs for each building, as well as determining when it
is more cost-effective to renovate a building with high maintenance expenses due to the frequent
occurrence of defects throughout the year.

2. Related works
Kwon et al. [11] propose a model that combines case-based reasoning and a genetic algorithm to predict
maintenance costs based on historical data. The approach involves retrieving similar cases from a case-
base and calculating case similarities based on their weights. In another study, Li and Guo [12]
conducted a study analyzing maintenance costs and budgets, considering variables such as building age,
number of floors, and elevators as independent variables and the accumulation of costs as the dependent
variable. Au-Yong et al. [13] identified five independent variables significantly correlated to
maintenance expenditure variance, including the skill and knowledge of managers, availability of
monitoring equipment and techniques, capability to adopt monitoring technology, and reliability of
maintenance data. Additionally, Kim et al. [9] conducted a study on maintenance and repair cost
estimation, considering factors such as building information, environmental vulnerability, and natural
hazards. Park et al. [14] focused on forecasting repair schedules in buildings based on input attribute
weights such as building coverage ratio, floor area ratio, number of buildings, number of floors, number
of households, management area, maintenance cost per unit area, and number of maintenance activities,
while computing similarity scores between existing and test cases.
In the present study, the annual costs of building corrective maintenance for each building are
predicted based on historical data of the number of past defects and building age. Initially, the future
number of building defects is estimated using these factors. Subsequently, the past maintenance costs
for the previous year(s) are considered, along with expected changes in taxes, materials, and workers'
wages costs. The influences of various factors that affect the cost of maintenance, such as building type,
area, location, weather conditions, number of floors, water supply system, Heating, Ventilation, and Air
Conditioning (HVAC), and other building services and users, have already been accounted for in the
analysis of past defects and their corresponding maintenance costs. The only remaining factors to
consider are the effects of increasing building age and changes in prices of taxes, materials, and workers'
wages, all of which have been considered in this study.

3. Methods
The present study aimed to analyze the historical data on building defects in 40 buildings with different
ages, sizes, and functions over a period of three consecutive years maintained mainly through a
corrective maintenance approach at Universiti Tun Hussein Onn Malaysia. These buildings included
offices, accommodations, laboratories, classrooms, and facilities. The defects data recorded all types of
building defects that occurred over the year in the buildings, which were mainly categorized into civil,
electrical, and mechanical. Also, information on each building, such as the number of defects, building
age, and building type. According to D’Ambro et al. [10], defect prediction based on past data of the

2
IConCEES-2023 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

defects is one of the most validation methods. In this study, by leveraging the linear relationship between
the number of building defects and the age of the building, a regression analysis was conducted—a
statistical methodology that examines the relationship between two or more variables, enabling the
prediction of an outcome variable [15]. It is commonly considered a very effective statistical method
that can be used both as an empirical and predictive technique to analyze the contribution of possible
new objects to the overall reliability of estimates [16]. Although Van and Morgan [17] have
recommended a minimum of 30 cases to ensure the regression model's ability to detect small effect
sizes, this study exceeded the minimum by including 40 buildings as cases. This ensured the robustness
and validity of the regression analysis. The regression model was constructed using numerical data on
defects of those buildings. Specific equations derived from the simple linear regression equation that are
used in this predictive regression model. As a result, this model and its equations were coded in Python
to evaluate their functionality and assure logical results. Running the equations on the dataset confirmed
that the model's predictions matched the expected outcomes.
Thus, this study used the Pearson product moment correlation coefficient (r) to analyze the
relationship between variables and the accuracy of model predictions to assure the model's accuracy and
validity. By using these rigorous measures, the validity of the predictive regression model and the
accuracy of its subsequent predictions can be assessed and verified.

4. Model development
Building’s age is a fundamental factor that affects the building’s life cycle in terms of cost, defects and
performance [18]. It plays a key role in the deterioration and occurrence of building defects [1]. As a
building ages, it requires additional maintenance and its value decreases. A building is supposed to start
with zero defects at handover, however defects steadily accumulate over time. Therefore, a building's
age directly influences the number of defects it experiences, with the number of defects being directly
proportional to the building's age. Numerous studies have confirmed that the age of a building affects
its life cycle, as it undergoes physical deterioration and may become obsolete, leading to a loss of
functionality. Figure 1 presents a line plot displaying the annual defect data for 40 buildings over a three-
consecutive-year span. The plot illustrates the relationship between the building age (x-axis) and the
number of defects (y-axis). Each building is represented by a line connecting the data points, with
building names annotated near the final point of each line. The inclusion of an overall regression line
provides an average trend for the relationship between age and defects.

Figure 1. Linear regression between the number of defects and building ages

Therefore, due to the presence of a linear relationship between building age as the predictor variable and
the total annual number of building defects as the response variable or the outcome variable, in

3
IConCEES-2023 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

accordance with the formulation by Kutner et al. [15] of the regression model when there is one predictor
variable, and the regression function is linear, as represented by Equation (1).

Y = β0 + β1X + εi (1)

Y represents the value of the outcome or response variable, while β0 is the y-intercept of the regression
line, providing the mean of the probability distribution of Y at X. The slope of the regression line, β1,
indicates the change in the mean of the probability distribution of Y per unit increase in X. X denotes
the predictor variable, and εi represents a random error term with a mean of 0. Therefore, the prediction
model of this study formulated as follows:

Predicted outcome = intercept + (slope × predictor value)

y ′ = a′ + b′ x ′ (2)

n ∑ xy−∑ x ∑ y
b′ = 2 (3)
n ∑ x2 −(∑ x)

a′ = y̅ − b′x̅ (4)

y′ = a′ + b′ * Building Age (5)

The prediction model used in this study is represented by the equation y′ = a′ + b′ x′, where y′ is the
total annual prediction of building defects (predicted outcome), x′ is the building age predictor, a′ is the
intercept, and b′ is the slope. To find out the regression coefficients, the model depends on the previous
annual defects (y) and the corresponding building ages (x) as input data. Also, y̅ represents the average
of y, while x̅ represents the average of x, and n represents the number of previous years. Hence, the total
annual prediction for the cost of corrective maintenance for each building can be obtained based on the
increment rate of the building defects of the predicted year. This coefficient reflects the projected rise
in repair costs as the number of defects in the building increases with its age. The total annual prediction
for the cost of defects can be calculated by Equation (6).

𝐶𝑚𝑐′= 𝐶𝑚𝑐 + (𝐶𝑚𝑐 * a) ± ctc (6)

To predict the total annual prediction for the cost of corrective maintenance for a specific building, by
using Equation (6). In this equation, 𝐶𝑚𝑐′ represents the predicted annual building corrective
maintenance cost. It includes the previous-year corrective maintenance cost of the building (𝐶𝑚𝑐), and
(a) is the ratio increment between the predicted total annual number of building defects (y′) and the total
annual number of building defects of the previous year (y), with (y') determined through Equation (5).
In other words, (a) represents the ratio increment or the percentage change in the number of building
defects from the previous year to the predicted year. And (ctc ) is represents the changes in costs
associated with taxes, materials, and workers' wages. The ± symbol represents the option to adjust the
cost upwards or downwards of taxes, materials, and workers' wages based on the latest rates; this
flexibility enables the inclusion of other relevant variables that may influence the overall cost estimation.
Thus, equation (6) can be used to predict the overall annual cost of corrective maintenance for a
particular building, taking into account its previous maintenance costs, the predicted rise in defects, and
any changes in costs linked to taxes, materials, and workers' wages.

5. Validation of model
In accordance with recent advancements in data collection, data analysis and computing resources,
predictive models have increasingly been used to generate predictions throughout multiple disciplines.
Thus, validation of the predictive models is critical to ensure the accuracy and quality of resultant
predictions. In the validation process, measures such as the Pearson correlation coefficient and model

4
IConCEES-2023 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

accuracy are typically used. The correlation coefficient assesses the relationship between variables as
well as the direction of the relationship. Therefore, this study has employed correlation coefficient and
model accuracy to assess the performance of the predictive model to ensure its reliability and validity,
enhancing its applicability in real-world use. The outcomes of this validation are presented as follows:

5.1 Pearson product-moment correlation coefficient

Table 1 presents the correlation between the number of building defects over three consecutive years.
The interpretation of the Pearson product-moment correlation coefficient value, which falls between
0.90 and 1.00, indicates a very high correlation [19].

Table 1. Correlation coefficient for number of building defects over three consecutive years
Number of Number of Number of
building defects building defects building defects
Data measurements year 1 year 2 year 3
Number of building Pearson Correlation 1 .995** .989**
defects year 1 Sig. (2-tailed) .000 .000
N 40 40 40
Number of building Pearson Correlation .995** 1 .995**
defects year 2 Sig. (2-tailed) .000 .000
N 40 40 40
Number of building Pearson Correlation .989** .995** 1
defects year 3 Sig. (2-tailed) .000 .000
N 40 40 40
**. Correlation is significant at the 0.01 level (2-tailed).

5.2 Model prediction accuracy

Furthermore, this study assessed the accuracy of the prediction model for the number of building defects
by predicting the number of building defects of the 40 buildings at year 3 as testing, based on their past
number of defects and ages from the previous two years. Then comparing their predicted number of
building defects at year three with their actual values by using the prediction model as in equation (5)
to make these predictions. Figure 2 illustrates that the prediction model achieved an accuracy rate of
96.06% in predicting the number of building defects. This high accuracy demonstrates the effectiveness
of the model in estimating the number of defects and reinforces its potential for practical application in
maintenance planning and decision-making processes.

Figure 2. Model prediction accuracy

5
IConCEES-2023 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

6. Implication and Practices

Currently, more demand is being applied to the building maintenance industry to estimate future costs
associated with building maintenance costs, mainly in public building organizations, for budget
allocation purposes. The prediction model of this study can estimate the ideal output of the number of
building defects and their maintenance cost for the next few years by inserting the input values to predict
the outcomes. Further, the prediction model can provide the right decision when the aged building is
better to be renovated rather than the choice of corrective maintenance approach for repairing their
defects based on the previous data for the previous years. This study established that the building
maintenance organizations that already have a system to archive the data of building defects and their
costs can predict next year maintenance costs in each building by adopting the prediction model that
was developed in this study.

7. Conclusions
This study introduces a novel predictive regression model for estimating the costs associated with
corrective maintenance for each building, facilitating informed decision-making regarding maintenance
strategies and building renovations. The predictive model has been validated to ensure logical outputs
and alignment with expected outcomes. The outcomes of this study have significant implications for
enhancing maintenance cost planning and budgeting, enhancing the accuracy of multi-year expenditure
projections. Moreover, the model assists in determining whether it is more advantageous to renovate
aging buildings or focus on addressing specific defects through corrective measures. In summary, the
developed model offers considerable benefits to building maintenance stakeholders, including
maintenance organizations, contractors, and clients, providing cost-effective decision-making, and
improving maintenance planning processes.
In the context of study limitations, it is important to note that the developed predictive model can
accurately forecast the annual cost of corrective building maintenance for the following year, assuming
the building continues to be maintained under the corrective maintenance policy and its function remains
unchanged. Although, the key predictors identified in this study, namely the number of past building
defects, building age, and previous maintenance costs, have been found to be effective in capturing and
reflecting other significant factors such as building type, size, levels, and users. However, it is crucial to
acknowledge that there may be limitations to the model's predictive capability in certain scenarios.
Additional adjustments might be required for further analysis, mainly when the corrective maintenance
approach is not consistently employed or if there have been changes to the building's function. In such
cases, the model's accuracy and applicability could be affected, and further considerations or
modifications may be necessary to enhance its performance.
To indicate the direction for future research, further study can train the model with maintenance cost
data from past years, as well as historical building defect data for buildings, and compare the actual and
expected costs of building corrective maintenance in each building.

8. References

[1] Hauashdh A, Jailani J, Abdul Rahman I and Al-Fadhali N 2022 Factors affecting the number of
building defects and the approaches to reduce their negative impacts in Malaysian public
universities’ buildings J. Facil. Manag. 20 145–71
[2] Chen C and Tang L 2019 BIM-based integrated management workflow design for schedule and
cost planning of building fabric maintenance Autom. Constr. 107 102944
[3] Mohd-Noor N, Hamid M Y, Abdul-Ghani A A and Haron S N 2011 Building Maintenance
Budget Determination: An Exploration Study in the Malaysia Government Practice Procedia
Eng. 20 435–44
[4] Alashari M, El-Rayes K, Attalla M and Al-Ghzawi M 2022 Multivariate time series and
regression models for forecasting annual maintenance costs of EPDM roofing systems J. Build.
Eng.
[5] Hauashdh A, Jailani J, Abdul Rahman I and AL-fadhali N 2020 Building maintenance practices

6
IConCEES-2023 IOP Publishing
IOP Conf. Series: Earth and Environmental Science 1347 (2024) 012025 doi:10.1088/1755-1315/1347/1/012025

in Malaysia: a systematic review of issues, effects and the way forward Int. J. Build. Pathol.
Adapt. 38 653–72
[6] Hauashdh A, Jailani J, Rahman I A and AL-fadhali N 2021 Structural equation model for
assessing factors affecting building maintenance success J. Build. Eng. 44 102680
[7] Ahzahar N, Karim N A, Hassan S H and Eman J 2011 A study of contribution factors to building
failures and defects in construction industry Procedia Engineering
[8] Hauashdh A, Jailani J, Rahman I A and AL-fadhali N 2022 Strategic approaches towards
achieving sustainable and effective building maintenance practices in maintenance-managed
buildings: A combination of expert interviews and a literature review J. Build. Eng. 45 103490
[9] Kim J M, Kim T, Yu Y J and Son K 2018 Development of a maintenance and repair cost
estimation model for educational buildings using regression analysis J. Asian Archit. Build. Eng.
[10] D’Ambros M, Lanza M and Robbes R 2012 Evaluating defect prediction approaches: a
benchmark and an extensive comparison Empir. Softw. Eng. 17 531–77
[11] Kwon N, Song K, Ahn Y, Park M and Jang Y 2020 Maintenance cost prediction for aging
residential buildings based on case-based reasoning and genetic algorithm J. Build. Eng.
[12] Li C S and Guo S J 2012 Life cycle cost analysis of maintenance costs and budgets for university
buildings in Taiwan J. Asian Archit. Build. Eng. 11 87–94
[13] Au-Yong C P, Ali A S and Ahmad F 2014 Prediction cost maintenance model of office building
based on condition-based maintenance Eksploat. i Niezawodn. 16 319–24
[14] Park S, Kwon N and Ahn Y 2019 Forecasting Repair Schedule for Building Components Based
on Case-Based Reasoning and Fuzzy-AHP Sustainability 11 7181
[15] Kutner M, Nachtsheim C and Neter J 2004 Applied linear regression models- 4th edition
[16] Kim G H, An S H and Kang K I 2004 Comparison of construction cost estimating models based
on regression analysis, neural networks, and case-based reasoning Build. Environ.
[17] Morgan B L and Voorhis C R W Van 2007 Understanding Power and Rules of Thumb for
Determining Sample Sizes Tutor. Quant. Methods Psychol.
[18] Yacob S, Ali A S and Au-Yong C P 2019 Establishing relationship between factors affecting
building defects and building condition J. Surv. Constr. Prop. 10 31–41
[19] Asuero A G, Sayago A and González A G 2006 The Correlation Coefficient: An Overview Crit.
Rev. Anal. Chem. 36 41–59

Acknowledgments
The authors would like to thank the Faculty of Civil Engineering and Built Environment, Universiti Tun
Hussein Onn Malaysia, and the Research Management Centre (RMC) of Universiti Tun Hussein Onn
Malaysia for supporting this study.

7
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.