Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Available online at www.sciencedirect.com
Procedia Computer Science 00 (2022) 000–000
Procedia Computer Science 00 (2022) 000–000 www.elsevier.com/locate/procedia
ScienceDirect www.elsevier.com/locate/procedia

Procedia Computer Science 207 (2022) 4087–4096

26th International Conference on Knowledge-Based and Intelligent Information & Engineering

26th International Conference on Knowledge-Based and Intelligent Information & Engineering
Systems (KES 2022)
Systems (KES 2022)
Machine Learning for Environmental Life Cycle Costing
Machine Learning for Environmental Life Cycle Costing
Aleksandra Markowskaaa, Małgorzata Krzywonosaa, Maria Čuljakb* , Ewa Walaszczykaa,
Aleksandra Markowska a,, Małgorzata
Karolina Miałkowska Krzywonos , MariaaČuljak
Anna Chojnacka-Komorowska , Kamal ,Matouk
b*
Ewa Walaszczyk
a
, Michał ,
Karolina Miałkowska , Anna Chojnacka-Komorowska
a
Śnierzyńskia
a
, Kamal Matouk , Michał
a

Śnierzyńskia
a
Wroclaw University of Economics and Business, Komandorska 118/120, Wroclaw 53-345, Poland
ba
University of Rijeka, Faculty
Wroclaw University of Economics
of Economics and Business,
and Business, Ivana Filipovića
Komandorska 4, 5100053-345,
118/120, Wroclaw Rijeka, Poland
Croatia
b
University of Rijeka, Faculty of Economics and Business, Ivana Filipovića 4, 51000 Rijeka, Croatia

Abstract
Abstract
Sustainable development lead to more mature and responsible production patterns and consumption. Nowadays the sustainable
Sustainable
approach is development
necessary for lead to more
adopting mature and solutions.
environmental responsibleTheproduction
aim of thispatterns
study and
is toconsumption. Nowadays
give an overview the sustainable
of scientific research
approach
developedisfor necessary for adopting
Environmental environmental
Life Cycle Costing withsolutions.
MachineThe aim of Combination
Learning. this study is to
of give an overview
Life Cycle Costingofasscientific research
an economic tool
developed
together for the
with Environmental Life Cycle
Machine Learning Costing
has been withtoMachine
shown be mostLearning.
effective.Combination of Life Cycle
It has been confirmed that Costing
MachineasLearning
an economic
helpstool
for
together
more with the
accurate Machinefor
predictions Learning has beenLife
Environmental shown
Cycleto be most effective. It has been confirmed that Machine Learning helps for
Costing.
more accurate predictions for Environmental Life Cycle Costing.
© 2022 The Authors. Published by Elsevier B.V.
© 2022
This The
is an Authors.
open accessPublished by ELSEVIER
article under B.V. license (https://creativecommons.org/licenses/by-nc-nd/4.0)
the CC BY-NC-ND
© 2022
This is The
an Authors.
open access Published
article by
under ELSEVIER
the CC
Peer-review under responsibility of the scientific B.V.committee
BY-NC-ND license of
(https://creativecommons.org/licenses/by-nc-nd/4.0)
the 26th International Conference on Knowledge-Based and
Intelligent Information & Engineering Systems (KES 2022) (https://creativecommons.org/licenses/by-nc-nd/4.0)
This is an
Peer-review open access
under article under
responsibility of the CC BY-NC-ND
scientific committeelicense
of KES International
Peer-review
Keywords: under
Life Cycleresponsibility
Cost; Machineof the scientific
Learning; committee
Alternative of KES Life
Cost Prediciton; International
Cycle Assessment;
Keywords: Life Cycle Cost; Machine Learning; Alternative Cost Prediciton; Life Cycle Assessment;

1. Introduction
1. Introduction
Sustainable development calls for starting more mature and responsible production patterns and consumption.
Sustainable
Economic growthdevelopment calls
itself impacts forseveral
us in starting morealso
ways, mature and responsible
the wrong ones. Socialproduction patterns
disparities and and environmental
constant consumption.
Economic growth itself impacts us in several ways, also the wrong ones. Social disparities and constant environmental
degradation are the product of our legacy for the next generations. In recent years there has been a constant growth in
degradation arearticles
the number of the product of our legacy
on machine forfor
learning theenvironmental
next generations. In but
LCC, recent
stillyears there
a few. has been
Machine a constant
learning growth
is mainly in
used
the number of articles on machine learning for environmental LCC, but still a few. Machine learning is
for LCC as an economic approach that sums up" total costs of a product, process or activity discounted over its mainly used
for LCC as an economic approach that sums up" total costs of a product, process or activity discounted over its

*
Corresponding author. Tel.: +385915433717; Email address: [email protected]
*
Corresponding author. Tel.: +385915433717; Email address: [email protected]

1877-0509 © 2022 The Authors. Published by ELSEVIER B.V.

1877-0509
This © 2022
is an open Thearticle
access Authors. Published
under by ELSEVIER
the CC BY-NC-ND B.V.(https://creativecommons.org/licenses/by-nc-nd/4.0)
license
Peer-review under
This is an open responsibility
access of the scientific
article under CC BY-NC-NDcommittee of KES
license International
(https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of KES International
1877-0509 © 2022 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 26th International Conference on Knowledge-Based and Intelligent
Information & Engineering Systems (KES 2022)
10.1016/j.procs.2022.09.471
4088 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096
2 Author name / Procedia Computer Science 00 (2022) 000–000

lifetime." However, it is generally associated with costs rather than just environmental costs. An environmental cost
is a relatively new approach; therefore, it should be considered in the E-LCA approach (a time-tested assessment
technique that evaluates environmental performance throughout the life cycle of a product or from performing a
service). The problem is in the multidisciplinary nature of the research. The sustainable approach contains several
ways of adopting environmental solutions. One of them is Life Cycle Costing (LCC) as a sustainable economic tool
and machine learning (ML) tool. This solution will be discussed in the article. The study aimed to verify and organize
scientific research developed for Environmental LCC with ML. The content analysis was performed to find the answer
to research questions, which were formulated as follows:
1. Which phases of ELCC are supported by machine learning?
2. Which methods of machine learning are used for supporting ELCC?
3. What kind of datasets are the most used for supporting ELCC by machine learning?
4. What kind of frameworks are proposed for supporting ELCC by machine learning?

2. Material and methods

Articles published in English and Polish related to environmental LCC and LCA using the Machine Learning.
The searches occurred in databases such as Scopus, Google Scholar, Web of Science, Stackoverflow, and Research
Gate. We used keywords: Life Cycle Cost, LCC, LCA, Machine Learning, Alternative Cost Prediction, Machine
Learning for Alternative Cost Prediction. The study was performed accordingly to the systematic literature review
protocol presented in Table 1.

Table 1. Systematic Literature Review Protocol.

GOAL To verify and organize scientific research developed for Environmental LCC with ML?
CONTEXT/FOCUS Articles published in English and Polish related to environmental LCC and LCA using the Machine
Learning tech.
RESEARCH QUESTIONS 1. Which phases of ELCC are supported by machine learning?
2. Which machine learning methods are used for supporting ELCC?
3. What kind of datasets are the most used for supporting ELCC by machine learning?
4. What kind of frameworks are proposed for supporting ELCC by machine learning?
DATABASES Scopus, Google-Scholar, Web of Science, ScienceDirect, Research Gate
SEARCH KEYWORDS Life Cycle Cost, LCC, LCA, Machine Learning, Alternative Cost Prediction, Machine Learning for
Alternative Cost Prediction, Deep Learning
INCLUSION CRITERIA Articles are written in the English and Polish languages. Publication date: not older than 2005.
EXCLUSION CRITERIA Not written in English and Polish; full text not available; duplicates in databases.
QUALITY ASSESSMENT Scopus (19 articles ), Google-SCHOLAR (7); WEB OF SCIENCE (10) RESEARCH GATE (2)
STACKOVERFLOW (1)
CONTENT ANALYSIS 26 articles
As a result, a sample of 26 articles was obtained, and each paper's content was analyzed in the context of research
questions. The results of this analysis are presented below.

3. Background

The first mention of the Life Cycle Assessment technique dates back to 1963, when Harold Smith presented his
research on the production of various types of energy in selected chemical processes at the World Energy Conference.
The conducted analysis covered all processes from the extraction of mineral resources to the production of the final
product. Analyzes of this type were then used in creating ecological policy, which consisted, for example, of
promoting the appropriate type of packaging or pro-ecological behavior of the society.
 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096 4089
Author name / Procedia Computer Science 00 (2019) 000–000 3

The original model of ecological assessment presented by Smith differed in form from the modern LCA structure,
but subsequent studies contributed to the improvement and development of this technique. In the same year, Coca-
Cola commissioned a comparative analysis of the amount of energy used, materials, and waste generated during the
production and use of beverage containers [2].
The entire theoretical basis of the LCA technique was established at a conference in Vermont in 1990, where the
term Life Cycle Assessment (LCA) was officially introduced [12]. It was then decided that for each assessed product
(or process), it was necessary to quantify the consumption of materials and energy for its production and analyze the
service life and all other life cycle phases, including storage.
The primary assumption of LCA is to consider all factors affecting the environment related to a given product. In
the case of the LCA technique, a product can be any good or service, but the analysis also to the product system, i.e.,
the entire production process, the product use phase or waste disposal. If LCC (Life Cycle Cost) assessments are
included in the LCA, it is possible to combine ecological and economic aspects in the decision-making process.
Currently, the development of the LCA technique in the world is focused on the exchange of data on unit processes
in individual industries (including by creating focused on the essence of the cost accounting of the ELCC cycle. The
available literature shows that life cycle costing is broken down into specific classifications. It can be concluded that
there is a conventional (traditional, business), thematic platforms) and the possibility of including economic aspects
(Life Cycle Cost) and social conditions (Social LCA) in the scope of the analysis.
The LCC environmental and social approach was analyzed by Testa et al. [20]. Also, various methodologies for
assessing ELCC cost accounting have been developed [16, p. 23].
Nevertheless, Hendrikson et al. [9], who took up the issue of cost accounting assessments, assumed that it is a
summary of all costs related to the product life cycle. One or more entities operating in the presented cycle covered
these costs. For example, it can be supplier, producer, user, consumer, people involved, plus environmental costs.
It should be noted that the life cycle account is very often overlooked in standard cost calculations of broadly
understood environmental protection, including the so-called hidden costs [13, p. 22]. The overall framework of cycle
costing is based on the physical product life-cycle, which requires an analysis of five major steps. These stages include
research and development, production, use and maintenance, and management in the liquidation and disposal phase.
Environmental LCA was conceived as a technique to complement environmental analysis. Therefore, it can be
assumed that LCA is a standardized method of assessing the impact of a given product on the environment and the
use of its resources in the so-called entire life cycle [3, p. 6-9].
Environmental Life Cycle Costing is applicable in many areas, especially in industry, construction and other
services. Also, in various design works, the current trends require that design be carried out in an eco-efficient way,
with the least resources and assuming that large masses of society will be able to use it. At the present stage of
development, each entrepreneur, also designing his product, determines what raw or auxiliary materials will be used
for a given production.

4. Results and discussion

The LCA is crucially important in new building design and renovations/refurbishments. However, it is hard to
predict the real-life service costs of different buildings (performance gap, functioning beyond warranty). Gao et al.
[6] presented using machine-learning techniques based on historical data to predict the LCC of facilities. The method
chosen relied on such components of LCC as multiple machine learning methods were tested to determine the best
ones for developing the LCC regression model, including multi-linear regression, support vector machine (SVM), k-
Nearest Neighbors (kNN), decision trees, multi-output SVM, multi-output regression trees, and multilayer perceptron.
The overall effect showed the possibility of gaining knowledge by investigating the feasibility of accurate estimations
of facilities' life-cycle costs (LCCs) by implementing machine learning on historical data. In the analyzed sample,
there is only a literature review on machine learning in predicting initial construction costs, maintenance and service
costs, and material consumption [4]. Gao et al. [4] identified the need for future research to use machine learning to
4090 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096
4 Author name / Procedia Computer Science 00 (2022) 000–000

predict the total life cycle costs of a building and highlight the underlying relationships between the various
components of LCC (i.e., maintenance costs, upfront costs, electricity costs). The need to develop a generalizable
framework for creating machine learning models for the construction of LCC analysis was also indicated. It is also
necessary to improve the quality and availability of data with the help of IoT technology.
LCCA is an economic analysis of a structure, which comprises the initial construction costs and costs due to
inspection, operation, repair, maintenance, and failure over its lifetime [21].
Yi et al. [23] stated that snowplow trucks are essential in winter maintenance by loading, removing, and disposing
of snow. The performance of the snowplow fleet is assessed through the life-cycle cost analysis basis, against which
the total cost curve is compared. These costs consist of maintenance, equipment purchase, and depreciation costs. The
snowplow truck performance analyses are performed throughout the entire life-cycle cost assessment at the macro
level to determine the optimum life cycle for the entire truck fleet. Unfortunately, this often leads to a waste of
resources. Life Cycle Costing was performed to determine the optimal life cycle for budget allocation purposes.
However, this assumption does not consider changes that result from exogenous factors, i.e., work intensity and work
conditions. Unfortunately, such analysis does not identify the factors contributing to performance deterioration. The
problems were solved by Yi et al. [23] by predictive ML techniques in a data-driven fashion.
A popular ML technique, random forest (R.F.), was presented. Another ML technique used was the linear support
vector machine (SVM) for comparison and benchmarking. The quantitative performance of the snowplows was
predicted based on the rank of the primary repair times. Extensive data has required the development of valid and
robust machine learning models [1]. Gau et al. [7] stated that the evolving systems of the building, such as
Computerized Maintenance Management Systems(CMMS), Building Automation Systems(BAS), and Building
Energy Management Systems (BEMS), have many valuable data. However, they have not been used to develop LCC
machine learning models. Having this data in machine learning, LCC is better understood in the institution, which
allows it to overcome many barriers associated with current LCC analysis methods. However, more research is
necessary to elucidate how to utilize the existing data housed indivisible building systems for LCC machine learning
model development. Gao et al. [6] used machine learning on historical data to estimate LCC. For this purpose,
acquiring LCC-related data (i.e., construction and design costs, maintenance work orders, utility consumption) from
many building systems was proposed. Later, all components of the LCC were introduced from historical data using
machine learning, and a case study on a university campus was conducted to demonstrate the proposed LCCA
structure.
Gao et al. [5] tried to format and separate LCC-related data stored and generated by different building systems. The
authors used machine learning techniques to predict each of the LCC components of a new building during the
programming stage. They also developed machine learning models for facility LCC analysis [4]. Gao et al. [4]
proposed a structure for creating machine learning models for facility LCC analysis that consisted of four main
modules: (1) obtaining descriptive attributes, (2) obtaining the target attributes, (3) training machine learning models
(Linear Regression, SVM Regression, KNN Regression, Regression Tree, and Multilayer Perceptron), and (4)
selecting the most suitable one and evaluating the models. With the target attributes and descriptive attributes ready,
the later step was to train the machine learning models based on these data. These methods have been successful in
forecasting construction costs [6]. The method pool of training regression in this framework model for facility LCC
analysis is expandable.
More methods can be adopted and implemented within the framework by developing machine learning techniques
in predictive data analytics. A considerable amount of energy is consumed by buildings, which considerably impacts
the environment. Reducing energy consumption is very important because this can be accomplished by renovating
existing ones or improving the design of new buildings. What is more, Simulation-Based Multi-Objective
Optimization (SBMO) models are used for assessing different renovation scenarios and optimizing LCCs and Total
Energy Consumption. Sharif et al. [19] proposed a generative deep learning building energy model using Variational
Autoencoders (VAEs). They highlighted several disadvantages of SBMO models, mainly due to the many calculations
necessary for the evolution process. They pointed out that the number of simulations must also be increased for the
 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096 4091
Author name / Procedia Computer Science 00 (2019) 000–000 5

accuracy level to increase, leading to an extended computation time. The limitation of the BEMs is that both ATHENA
and DesignBuilder (software for simulation) do not consider every aspect of renovation projects. For example, the
impact of elements removed in the renovation process is not included in the calculations. Namely, the trained ANN
models were only suitable for similar buildings, as only one data set from the institutional building was used for ANN
training, testing, and validation. It follows that the ANN-trained models are appropriate only for similar buildings.
The cost prediction model for refinery equipment using Life Cycle Costing Analysis is presented by Radziah et al.
[17]. Risk analysis is essential for life cycle costing and maintaining all equipment. It is applied in the research to
ensure business companies such as the gas and oil industry optimize their financial expenses. They can replace the
risky equipment before unwanted cases happen and predict the occurrence of the risk. In the project, the authors tried
to understand life cycle costing and integrate the concept to mitigate the risk of failures in refinery equipment and
estimate the possible risks, calculate, sustain cost, and repair. The general approach to deciding the alternative is based
on costs' lowest net present value. LCC is the present value of cumulated future annual cost throughout the examined
timeframe. LCC are presented as general formula (1):

𝑇𝑇
𝐶𝐶𝑡𝑡
∑ (1+𝑟𝑟)𝑡𝑡
(1)
𝑡𝑡=0

where:
NPV - the present value of LCC,
Ct - the sum of every relevant cost after deducting yields generated during the period t,
r - the discount rate,
t - the analysed time (t=0…T) (years),
T - the life cycle.

From the synthetic data, LCC was computed using formula 2:

LCC=AC+IC+OC+LP+(RC−SC) (1+ Interest Rate) (Year−Base Year) (2)

where:
A.C. - Alternate Current Corrosion/Acquisition Cost,
I.C. - Internal Corrosion/Initial Cost,
O.C. - Operation Cost,
L.P. - Lost Production,
RC - Replacement/ Repair Cost,
SC - Residual Value.

Besides RBI, pre-processing or data preparation takes place in the LCC records, wherein the LCC column,
normalization is needed to have a proper comparison with a standard scale of 0 to 1 by using the formula 3:

𝑋𝑋𝑛𝑛𝑛𝑛𝑛𝑛 = 𝑋𝑋 − 𝑋𝑋𝑚𝑚𝑚𝑚𝑚𝑚 𝑋𝑋𝑚𝑚𝑚𝑚𝑚𝑚 − 𝑋𝑋𝑚𝑚𝑚𝑚𝑚𝑚 (3)

For further development needs, recommended associating the consequences of aging equipment on the
environment, such as burns, explosions, fatal accidents, and others which can give a clearer view for operators to take
explicit action on equipment that nears to disposal period. Furthermore, the forecasting of the inflation rate affects the
general LCC of equipment. Intelligent techniques to efficiently implement product configuration have been proposed
to enable mass customization. Based on existing product life-cycle data, a rule mining approach was applied to
4092 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096
6 Author name / Procedia Computer Science 00 (2022) 000–000

generate configuration knowledge automatically. This knowledge led to constraints between components within a
product family, which can facilitate a product configurator to improve its efficiency.
Machine learning techniques and the Activity-Based Costing (ABC) technique were applied for estimating product
LCC at the early design stage [15]. ABC systems provide knowledge about the costs of products and support and
production activities. Due to ABC understanding how costs are consumed, designers can develop economical designs.
Liu et al. [15] estimated the LCC of a new product variant of a family by implementing the following steps:
1) The life-cycle activities which incur the cost were identified.
2) Resource cost drivers were identified, and measured resource consumption rates. The consumption rate C.R. of a
resource was calculated according to the formula (4):

CR=total cost driver quantity (4)

The total resource costs consumed by activity with the known resource consumption rate were calculated using
formula (5):

AC=∑i=1mCRi×RUQiAC=∑i=1mCRi×RUQi

∑𝑚𝑚
𝑖𝑖=1 𝐶𝐶𝐶𝐶𝑖𝑖 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖 ∗ 𝐴𝐴𝐴𝐴 (5)

where:
m- the number of various resources consumed by the activity,
CRi - the consumption rate of the ith resource,
RUQi - the utilization quantity of the ith resource driver.

Next, an activity cost driver was used to measure how a cost was incurred for an activity. With the established total
cost activity cost AC, the activity unit cost rate was calculated as (6):

UR=ACAUQ (6)
where:
AUQ - the quantity of determined activity cost driver.
The authors calculated the (partial) LCC of a component/product based on formula (7):

∑𝑛𝑛𝑗𝑗=1 𝑈𝑈𝑈𝑈𝑗𝑗 ∗ 𝐶𝐶𝐶𝐶𝐶𝐶𝑗𝑗 (7)

n - the total number of involved life-cycle activities in the product,

URj - the unit cost rate of the jth activity,
CAQj - the estimated consumption quantity of the j th activity.
The authors chose to develop SVM and ANN and prediction models based on the steps: constructing data, setting
parameters, training machine, validating machine, predicting cost, and saving machine. Both models showed
consistent performance under various noise data and training data noise. Therefore, different models to predict the
LCCs of newly derived configurations were used, depending on the quality of training data available in PLD.
Life cycle cost (LCC) of equipment is an economic strategy derived from the International Organization for
Standardization (ISO) 15686, which was first proposed by the U.S. Department of Defence in the 1960s. LCC is
something necessary as quality of products and the functionality regarding the decision of product success in the here
and now. In contrast, other current cost estimating techniques show low accuracy and are limited to specific life-cycle
phases. It is explained as the total expenses spent on the equipment while design, development, production, and use
until a scrap of the product. The component parameters that make up the LCC depend on the industry. The usage costs
 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096 4093
Author name / Procedia Computer Science 00 (2019) 000–000 7

of a product in its life cycle are a few times more expensive than the procurement costs. So, the main goal of LCC
determination is to control and save costs. The result is to promote the reliability and maintainability of the equipment
and the maintenance of the equipment scientific management. The systems for LCC calculation can be used at the
very early stages of the product design process and, therefore, will provide information to designers on time. Currently,
used LCC methods are based on engineering estimation, parameter estimation, and expert estimate estimation. [2]
have observed that and concluded that the accrual estimation of the expenses aids the control of the expenses. The
estimation methods have some gaps, and there are some problems with them – in the example, because of too big a
workload or not factual accuracy. The solution for this could be the application of machine learning techniques and a
neural network model to estimate the expenses of the equipment. LCC is also applied as one component of the
equipment fleet management process regarding the equipment replacement optimization problem. Hejazian et al. [8]
presented the best time to replace and solve the mentioned optimization problem of contactors by the economic life
model and examined these machines' total cost. The proposed formula for LCC calculation in the case of forest
machines is presented below (8):
𝐿𝐿𝐿𝐿𝐿𝐿 = 𝐶𝐶𝐷𝐷 + 𝐶𝐶𝐼𝐼 + 𝐶𝐶𝐼𝐼𝐼𝐼 + 𝐶𝐶𝑇𝑇𝑇𝑇 + 𝐶𝐶𝑃𝑃 + 𝐶𝐶𝑀𝑀,𝑅𝑅 + 𝐶𝐶𝑇𝑇𝑇𝑇 + 𝐶𝐶𝐿𝐿𝐿𝐿 + 𝐶𝐶𝐹𝐹 + 𝐶𝐶𝑙𝑙𝑙𝑙 (8)
where:
𝐶𝐶𝐷𝐷 + 𝐶𝐶𝐼𝐼 + 𝐶𝐶𝐼𝐼𝐼𝐼 + 𝐶𝐶𝑇𝑇𝑇𝑇 + 𝐶𝐶𝑃𝑃 – ownership costs
𝐶𝐶𝐷𝐷 - depreciation cost,
𝐶𝐶𝐼𝐼 - interests cost,
𝐶𝐶𝐼𝐼𝐼𝐼 - insurance cost,
𝐶𝐶𝑇𝑇𝑇𝑇 - tax,
𝐶𝐶𝑃𝑃 - parking cost.
𝐶𝐶𝑀𝑀,𝑅𝑅 + 𝐶𝐶𝑇𝑇𝑇𝑇 + 𝐶𝐶𝐿𝐿𝐿𝐿 + 𝐶𝐶𝐹𝐹 + 𝐶𝐶𝑙𝑙𝑙𝑙 – operating costs
𝐶𝐶𝑀𝑀,𝑅𝑅 - maintenance and repair costs,
𝐶𝐶𝑇𝑇𝑇𝑇 - tire cost,
𝐶𝐶𝐿𝐿𝐿𝐿 - lubricant cost,
𝐶𝐶𝐹𝐹 - fuel cost,
𝐶𝐶𝑙𝑙𝑙𝑙 - labor cost.

The main advantage of currently applied LCC methods is the strong adaptive learning ability. The regression
modeling is becoming more accurate. Moreover, the prediction accuracy can meet the requirements of the engineering
application. Nowadays, some solutions regarding LCC have been developed. Liu et al. [14] proposed to build up an
efficient and intelligent LCC estimation system that overcomes the drawbacks of existing systems. Its main concept
is a generic system that allows the users to enter the ABC technique and ML techniques to define and estimate various
LCC elements. Those elements depend on the information available in a Product Life-Cycle Database (PLD). The
main concept focuses on five major components: PLD, LCC template manager, ABC module, ML module, and
synthesizer. The following solution where LCC estimation is possible is the application of the Back Propagation (BP)
learning algorithm [10]. This algorithm can be used to have the memory of the training and, as a result, will be able
to construct the network model and set up the relationship between the parameters and cost. The influence factors of
the cost were the neural networks input data, and life-cycle cost data was the prediction output. The proposed structure
consists of a three-layer that has single output and multi-input. During the training process, matrix operations have
been applied. If the calculated error reached the set value, the calculation stopped, generated network, and called the
network to forecast. The MATLAB environment has been used to build the model. Besides the solution mentioned
above, the least square support vector machine (LSSVM) model based on genetic algorithm has also been proposed
for LCC prediction [2]. LSSVM is an improvement of SVM. It uses the square sum error loss function to replace the
insensitive loss function in SVM. Then it applies the equality constraint to obtain the linear equations (9):
𝐷𝐷 = {(𝑥𝑥𝑖𝑖 , 𝑦𝑦𝑖𝑖 )|, 𝑖𝑖 = 1, 2, 3, … , 𝑁𝑁} (9)
4094 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096
8 Author name / Procedia Computer Science 00 (2022) 000–000

where:
D – sample set,
𝑥𝑥𝑖𝑖 - input sample,
𝑦𝑦𝑖𝑖 - output sample.
𝑦𝑦(𝑥𝑥) = 𝜔𝜔𝑇𝑇 𝜑𝜑(𝑥𝑥𝑖𝑖 ) + 𝑏𝑏 + 𝑒𝑒𝑖𝑖 (10)
where:
y(x) – model of sample data,
𝜔𝜔- weight vector,
𝜑𝜑(𝑥𝑥𝑖𝑖 ) - non-linear function,
b – deviation amount,
𝑒𝑒𝑖𝑖 – fitting error.
Gao et al. [5] proposed a probabilistic LCC prediction model developed by applying the Monte Carlo simulation
(MCS) to an LCC prediction model based on case-based reasoning (CBR) to grant the possibility to support the process
of decision making in the early stages of construction projects.
In extension to the work of Gao et al. [4], Gao et al. [6] presented the application of Advances Particle Swarm
Optimization (APSO) Optimized LSSVM Algorithm. The parameters of the LSSVM model have been optimized
before is possible to avoid the problem that the parameters of the traditional LSSVM model depend on experience,
and the accuracy of simulation results is low. The prediction and performance indexes of four applied models: APSO
LSSVM, traditional LSSVM, and BP neural network, have been analyzed and compared. As a result, the APSO
LSSVM prediction model has better prediction accuracy than other models. LCC based on the LSSVM regression
model has been calculated as follows (11) (eliminating 𝜔𝜔 𝑎𝑎𝑎𝑎𝑎𝑎 𝑒𝑒𝑖𝑖 ):
𝐿𝐿𝐿𝐿𝐿𝐿(𝑥𝑥) = ∑𝑁𝑁
𝑖𝑖=1 𝛼𝛼𝑖𝑖 𝐾𝐾(𝑥𝑥, 𝑥𝑥𝑖𝑖 ) + 𝑏𝑏 (11)
where:
𝑥𝑥𝑖𝑖 - the input variable and the total cost of LCC is the output of the prediction model,
𝐾𝐾(𝑥𝑥𝑖𝑖 , 𝑥𝑥) – Kernel,
α and b - solution of formulas.
As one of the top contributors to global warming and pollution, agriculture is an area of interest in quantifying life
cycle environmental impact to form possible prevention actions. Romeiko et al. [18] determined the most efficient
and accurate model for estimating spatially explicit life cycle global warming and eutrophication impacts at the county
scale, with corn production in the US's Midwest region as a case study. This study compared the predictive accuracies
and efficiencies of five distinct supervised machine learning (ML) algorithms. ML algorithms are one million times
faster than the traditional process-based model with high predictive accuracy. ML can serve as an alternative surrogate
of process-based models to estimate life-cycle environmental impacts, capturing large geographic areas and
timeframes. This study found that GBRT and XGBoostmodels yielded the highest predictive accuracy for both the
life cycle of global warming and eutrophication. The study was primarily scaled with data gathered from over 1000
counties in 9 years. Authors [18] recommended future studies to test methods used in this research for predicting
spatially explicit life cycle environmental impacts of soybean, beef, and dairy production, which are among the top
contributors to the national global warming and eutrophication impacts. ML models give new opportunities for solving
computation-intensive challenges, such as optimization, prediction, validation, and supply chain. It has been shown
that the ML tools help for more accurate prediction of LCC facilities. Especially SVM and RF techniques improve
analysis to identify the factors contributing to performance deterioration. The Variational Autoencoders were
successful considering computational accuracy and time. ML techniques lead to breakthroughs in energy prediction
in construction. An essential factor in ML algorithms is the data's starting point. More data positively affect the
generalization and accuracy in predicting the generative models. Necessary steps in using ML tools are constructing
data, setting parameters, training the model, validating the model, and predicting using the model.
The study of LCC/LCA itself has its boundaries. The goal is to develop machine learning systems that gather
comprehensive data like the economy, geodata, or natural sciences. Development of this kind of model may provide
 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096 4095
Author name / Procedia Computer Science 00 (2019) 000–000 9

high-quality data to help resolve real-life problems, lead to a real, sustainable economy, and reduce the adverse impact
on nature.

5. Conclusions

The general conclusion to this literature overview opens a field for research in Machine Learning in environmental
LCC, although there is an interest and optimism among scientists. The few research and development studies show
that ML is a powerful tool in high-quality prediction building models for LCC and LCA. Machine learning is used
mainly for LCC as an economic approach that sums up the total costs of a product, process, or activity discounted
over its lifetime. It is, however, associated with cost in general, rather than just environmental costs. An environmental
cost is a new approach; therefore, it should be considered in the E-LCA approach (a time-tested assessment technique
that evaluates environmental performance throughout the life cycle of a product or from performing a service). The
problem is in the multidisciplinary nature of the research.
Acknowledgements
The project is financed by the Ministry of Science and Higher Education in Poland under the programme "Regional
Initiative of Excellence" 2019 - 2022 project number 015/RID/2018/19 total funding amount 10 721 040,00 PLN.

References

[1] Alpaydin E. (2014) "Introduction to machine learning", Cambridge, 3rd ed, MA: MIT Press.
[2] Chen Y.H, SU Y.S. and DU X.H. (2020) "Research on Equipment Life Cycle Cost Prediction Based on GA-LSSVM", Journal of Physics:
Conference Series, [in:] https://doi.org/10.1088/1742-6596/1605/1/012089.
[3] Ciambrone F.D. (1997) "Environmental Life Cycle Analysis", Lewis Publisher, New York.
[4] Gao X, Pishdad-Bozorgi P., Shelden D.R., and Hu Y. (2019) "Machine Learning Applications in Facility Life-Cycle Cost Analysis: A Review",
Computing in Civil Engineering.
[5] Gao X., Pishdad-Bozorgi P., Shelden D.R. and Tang S. (2020) "Machine Learning-Based Building Life-Cycle Cost Prediction: A Framework
and Ontology", Construction Research Congress 2020.
[6] Gao X., Pishdad-Bozorgi P., Shelden D.R., and Tang S. (2019) "Machine Learning-based Lifecycle Cost Analysis for Educational Facilities".
Georgia Institute of Technology Atlanta, GA, USA.
[7] Gau X. and Pishdad-Bozorgi P. (2019) "A framework of developing machine learning models for facility lifecycle cost analysis".
[8] Hejazian M., Lotfalian M. and Limaei S.M (2018) "Estimating the economic life of forest machinery using the cumulative cost model and cost
minimization model in Iranian Caspian forests", Journal of Forest Science, [in:] https://doi.org/10.17221/133/2017-JFS.
[9] Hendrikson CT. and Lave LB. and Matthews H.S. (2010) "Matthews, Environmental Life Cycle Assessment of Goods and Services. An Input
and Output Approach RFF", Washington.
[10] Jin J., Geng J., Sun L. and Chen G. (2012) "Life cycle cost estimated for the diesel in mechanical engineering based on the BP neural network",
Lecture Notes in Electrical Engineering, [in:] https://doi.org/10.1007/978-3-642-31507-7_34.
[11] Jin Z., Kim J., Hyun C.T. and Han S. (2019) "Development of a Model for Predicting Probabilistic Lifecycle Cost for the Early Stage of Public-
Office Construction", Sustainability, [in:] https://doi.org/10.3390/su11143828. [5]
[12] Kędzierska E. (2004) “Analiza LCA (Life Cycle Assessment – ekologiczna ocena cyklu życia) procesów przetwórczych w ZG Trzebionka
S.A.”, Praca Magisterska, Instytut Chemii i Technologii Nieorganicznej Politechniki Krakowskiej, Kraków.
[13] Lewandowska A., Foltynowicz Z. and Podleśny A. (2004) “Środowiskowa Ocena Cyklu Życia (LCA) – zastosowania, Problemy Ekologii”,
Vol. 8, nr 3.
[14] Liu H., Gopalkrishnan V., Ng. W.K., Song B. and Li X. (2008) "An intelligent system for estimating full product life cycle cost at the early
design stage", International Journal of Product Lifecycle Management, [in:] https://doi.org/10.1504/IJPLM.2008.021436.
[15] Liu H., Huang Y., Ng W.K., Song B., Li X. and Wen-Feng Lu (2007) "Machine learning techniques and the ABC technique were applied for
estimating product LCC at the early design stage".
[16] Nilsson M. and Eckerberg K. (2009) "Environmental policy integration in practice: Shaping institutions for learning", Earthscan, London.
[17] Radziah Z., Jafreezal J., Izzatdin A.A. (2018) "Risk Overview and Cost Prediction Model for Refinery Equipment Using Life Cycle Costing
Analysis", 2018 IEEE Conference on Big Data and Analytics (ICBDA).
[18] Romeiko X.X., Guo Z., Pang Y., Lee E.K. and Zhang X. (2020) "Comparing Machine Learning Approaches for Predicting Spatially Explicit
Life Cycle Global Warming and Eutrophication Impacts from Corn Production ", [in:] https://www.mdpi.com/2071-1050/12/4/1481/htm
4096 Aleksandra Markowska et al. / Procedia Computer Science 207 (2022) 4087–4096
10 Author name / Procedia Computer Science 00 (2022) 000–000

[19] Sharif S.A., Hammad A., Eshraghi P. (2020) "Generation of Whole Building Renovation Scenarios Using Variational Autoencoders", Journal
Pre-proofs.
[20] Testa F., Iraldo F., Frey M. and Connor R.O. (2020) "Life Cycle Costing, a View of Potential Applications: from Cost Management Tool to
Eco-Efficiency Measurement", [in:] www.intechopen.com, data wejścia 1.03.2020.
[21] Valigura J., Lieb A.B. and Sideris P. (2020) "Lifecycle cost assessment of conventional and hybrid sliding-rocking bridges in seismic areas",
Structure and Infrastructure Engineering, DOI: 10.1080/15732479.2020.1766514.
[22] Yi X., Chuanyu X., Wu W., Zhiwei X., Lie L., Xiaohong L., Lipin S., Qiupeng Z., Yuxin Z., Shuang L., Yue M., Ting G., Li M. (2019) "Life
Cycle Cost Prediction of Substation Based on Advanced PSO and Least Squares Support Vector Machine", 5th International Conference on
Energy Equipment Science and Engineering, [in:] https://doi.org/10.1088/1755-1315/461/1/012083. [6]
[23] Yi Z., Liu X.C., Wei R., Grubesic T.H. (2021) "Snowplow Truck Performance Assessment and Feature Importance Analysis Using Machine-
Learning Techniques", [in:] https://ascelibrary.org/doi/epdf/10.1061/JTEPBS.0000486.