https://doi.org/10.31449/inf.v48i13.6013 Informatica 48 (2024) 81–96 81 
 
Construction of Sustainable Building Performance Optimization 
Design Model Based on Sensitive Multi-objective Decision-making 
 
Qinhui Fang
1*
, Jing Zhang
2
, Ze Xu
1
, Yaou Lv
1
, Jian Pi
1
 
1
HCIG Xiong'an Construction Development Co., Ltd, Xiongan, 070001, China 
2
HCIG Operation Management Co., Ltd, Xiongan, 070001, China 
E-mail: fangqinhui1971@126.com 
Keywords: sensitivity, multi-criteria decision-making, sustainable building, genetic algorithm, glass heat transfer 
coefficient 
Received: April 12, 2024 
In view of the frequent changes in the building performance scheme and the many factors affecting the 
indicators, it is difficult to achieve better application accuracy and optimization effect only based on 
the experience of architects. Moreover, due to the limitation of professional ability, it is difficult to 
analyze the architectural design process directly with the help of intelligent algorithms, and the 
interactivity is poor. Based on this, the study proposes an optimization design scheme based on 
sensitivity multi-objective decision-making, which takes the sensitivity factor into consideration, 
optimizes the design scheme based on the relationship between architectural variables, and achieves 
the calculation of the fitness function of the multi-objective problem with the help of intelligent 
optimization algorithms. The results indicated that the improved algorithm has better convergence and 
effectiveness when carrying out the selection of program indicators, and it can obtain the average 
number of performance-attained design solutions is 5.58. After the improvement of the building 
program, it was found that the influence weight of the glass heat transfer coefficient accounted for a 
larger proportion, and the error value of the indicator results was smaller, which effectively improved 
the optimization efficiency of the building. In addition, in the optimized design of the scheme, the 
number of modifications were less than 12 times, which was much smaller than the number before 
optimization. The proposed method can effectively provide reference value for architects to carry out 
program optimization. 
Povzetek: Prispevek predlaga model za optimizacijo trajnostne gradbene zmogljivosti, ki temelji na 
občutljivem večciljnem odločanju. Uporaba inteligentnih algoritmov izboljša natančnost in 
učinkovitost zasnove ter zmanjša število potrebnih sprememb.
1 Introduction 
The development of the global concept of sustainable 
development and the worsening energy crisis have made 
sustainable building (SB) has gradually become the 
mainstream selection. SB emphasizes on minimizing the 
negative impact on the environment during the whole 
cycle of building design, construction, operation and 
demolition, while ensuring the building function, comfort 
and economy, which has the characteristics and 
advantages of green environment protection, energy 
saving and emission reduction [1]. Existing SB design 
methods have made some progress in improving energy 
efficiency, utilizing renewable energy, and reducing 
resource consumption, but there are still limitations. For 
example, traditional methods often do not consider 
comprehensively enough when dealing with the 
optimization of multiple aspects of building performance 
(BP), or are inefficient in achieving the optimal design 
solution, and the multi-objective nature of BP 
optimization design makes the design process complex 
and difficult to achieve the optimal balance of various 
indicators [2]. In the construction process, the impact of 
the physical and chemical environments on the building 
should be taken into account, analyzed comprehensively, 
and the relationship between different factors should be 
grasped in order to effectively realize the embodiment of 
BP advantages [3]. Therefore, the study proposes a 
sensitivity-based multi-criteria decision-making 
algorithm (SMDA) for the SB design model in order to 
ensure the balance between multiple performance 
objectives, aiming to provide building designers with an 
effective decision support tool. The sensitivity 
measurements of the relevant parameters of the design 
solution are generally measured mostly by direct or 
indirect methods, and the specific methods involved are 
local sensitivity analysis (LSA), labeled regression 
coefficient ordering, Fourier amplitude sensitivity 
experiments, and Sobol global sensitivity analysis. LSA 
is computationally simple, easy to understand and can be 
applied to most evaluation models. The innovation of the 
study is to identify and quantify the key factors affecting 
BP through sensitivity analysis, and the application of 

82   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
multi-criteria decision-making (MCDM) method and 
improved algorithms can also be used to optimize the 
building scheme by considering multiple performance 
indicators, thus ensuring the environmental and economic 
performance of the building. 
2 Related works 
MCDM is widely used in various industries as a scheme 
for evaluating and selecting decision objectives using 
multiple criteria. To extensively study the sensitivity of 
multi-criteria decision-making methods, scholars such as 
Nabavi and Wang proposed a method that uses eight 
multi-criteria decision-making methods fused with 
entropy methods and correlation between criteria. The 
experimental results indicated that for some of the 
multi-criteria decision-making methods such as gray 
relational analysis and example-based portfolio 
evaluation method and simple phase weighting method 
are not sensitive to three of the alternative 
decision-making schemes [4]. Scholars such as Basten 
developed a model for MCDM in order to minimize 
product maintenance time and repair cost. This model 
analyzed the total maintenance cost and total maintenance 
time after generating Pareto bounds using the ε
-constraint method [5]. To develop the ratio analysis of 
Pythagorean fuzzy multi-objective optimization with the 
addition of a full multiple-form approach to solve 
multi-criteria decision problems with completely 
unknown information about the criterion weights, Sarkar 
and Biswas devised a new distance metric model by 
combining the Hamming distance and the Hausdorff 
metric. The model verified the sensitivity of the proposed 
model by varying the criterion weights affecting the 
strategy hierarchy [6]. In the field of disposable device 
technology for biopharmaceuticals, Zurcher et al. 
proposed a multi-stage decision support method with the 
objective of reducing the process risk in the 
pharmaceutical process. This method was experimentally 
verified by the researchers, who systematically explored 
and researched the uncertainty in each stage of the 
pharmaceutical process using the proposed method [7]. In 
order to study the safe water, use in various places under 
the global warming environment, Deshbhushan and Rajiv 
proposed a centralized rainwater harvesting model 
integrating GIS and multi-criteria decision making. The 
study compared various predefined criteria by using 
criterion importance and entropy of inter-criteria 
correlation and verified in subsequent experiments that 
the proposed model is informative for practical 
application assessment [8]. 
In terms of BP, a large number of current examples 
show that multi-objective optimization methods can 
significantly improve BP. To enhance the interaction 
between the optimization process and the architectural 
process, Lin et al. proposed a preference-based MCDM 
method to realize the designer's decision-making 
preferences during the optimization process, which makes 
the objective more reliable and improves the optimization 
efficiency at the same time. Experimental results 
demonstrated that the constructed method has better 
convergence and higher preference satisfaction, which is 
more applicable to the objective decision-making process 
in the early design stage [9]. To ensure the coexistence 
between different analysis methods used in the building 
process, Utkucu and Sozer proposed an approach using 
building information modeling and integrating the 
MCDM method and multi-criteria eugenic evaluation. 
Experimental data demonstrated that the proposed 
method is able to achieve energy savings of more than 
75% and accurately categorize the energy performance of 
different buildings based on a detailed energy analysis 
[10]. Xie and Wang used a hybrid MCDM approach to 
propose a multi-objective lightweighting and 
crashworthiness optimization design method for complex 
cross-sectional shapes and sizes of S-rails to determine 
the set of optimal solutions to weigh the optimal points. 
The experimental results showed that the total mass of the 
optimized model is reduced by 25.62% and the 
performance is improved compared to the initial 
optimized model [11]. In terms of the configuration of 
buildings, Seyed et al. proposed a robust clustering 
multi-objective method based on MCDM to address the 
uncertainty in the input variables of the pipeline 
maintenance distribution network. This method 
incorporated hydraulic simulation, pipeline failure rate 
prediction, nonlinear interval planning, and multi-criteria 
decision making. It was important to note that pipeline 
maintenance is costly and must be carried out on a 
regular basis. The method was experimentally 
demonstrated to be able to improve the physical 
performance of the network by 56% and the hydraulic 
performance by 35% when implementing optimal 
instructions, and reduce the annual deficit in node 
demand to 69% [12]. The study organized the content of 
past literature and obtained Table 1. 
 
Table 1: Compilation of past literature ideas 
References Methods Key findings Limitations 
[4] 
Using eight multi-criteria 
decision-making methods to 
integrate entropy method and inter 
criteria correlation for scheme 
selection 
Grey relational analysis and 
case analysis show poor 
sensitivity 
Lack of further definition of 
the application of limited 
methods in decision-making 
problems of different natures 
[5] 
 
Established a multi-criteria decision 
management (MCDM) model for 
This method can reduce 
maintenance time to a certain 
The application scope of the 
model may be limited to 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 83 
 
products to ε- constraint method for 
generating Pareto boundaries 
extent specific maintenance 
optimization problems, and 
constraint methods may be 
difficult to consider the 
multidimensional factors of 
product management 
[6] 
A metric model combining 
Hamming distance and Hausdorff 
metric to solve multi criteria 
decision-making problems 
This model verifies the 
sensitivity of the proposed 
model by changing the 
standard weights that affect 
the policy hierarchy 
The complexity of the model 
may face difficulties in 
determining parameters in 
practical applications, and 
weight adjustments may 
overlook the consideration of 
expert knowledge and 
experience 
[7] 
Research on technical issues of 
disposable biopharmaceutical 
equipment using multi stage 
decision support method 
Reduced pharmaceutical 
process risks and validated 
the uncertainty of each stage 
in the pharmaceutical process 
through experiments 
This technology needs to be 
validated in different 
environments and technologies 
[8] 
A centralized rainwater collection 
model integrating geographic 
information systems and multi 
standard decision-making was 
proposed, which compares 
predefined standards based on the 
importance of standards and the 
correlation entropy between 
standards 
This method has reference 
value for the practical 
application evaluation of safe 
water use 
The applicability and accuracy 
of the model may be limited by 
data quality and availability. 
[9] 
A preference based architectural 
MCDM method was proposed to 
optimize the decision preferences of 
designers 
This method has better 
convergence and preference 
satisfaction, improved 
optimization efficiency, and is 
more suitable for the target 
decision-making process in 
the early design stage 
There may be an excessive 
reliance on designer preference 
information 
[10] 
Integrating MCDM method and 
multi standard eugenics evaluation 
method through building 
information modeling 
This method can achieve 
energy-saving effects of over 
75% and can be classified 
based on detailed energy 
analysis 
The complexity of the method 
is high, and the scope of 
promotion and application will 
be limited to a certain extent 
[11] 
A multi-objective lightweight and 
crashworthiness optimization 
design method is proposed for 
S-shaped guide rails using a hybrid 
MCDM method 
The total mass of the 
optimized model has been 
reduced by 25.62%, and the 
performance has been 
improved 
The specificity between 
determining the optimal 
solution and setting conditions 
makes the method difficult to 
generalize 
[12] 
A robust clustering multi-objective 
method based on MCDM was 
proposed to solve the uncertainty 
problem of input variables in 
pipeline maintenance distribution 
networks. 
The physical performance of 
the pipeline network has been 
improved by 56%, the 
hydraulic performance has 
been improved by 35%, and 
the annual deficit of node 
demand has been reduced to 
69% 
The specificity between 
determining the optimal 
solution and setting conditions 
makes the method difficult to 
generalize 
Research 
method 
Propose an optimization design 
scheme based on sensitivity 
multi-objective decision-making, 
and use intelligent optimization 
algorithms to calculate the fitness 
This improved algorithm has 
good convergence and 
effectiveness, with an average 
of 5.58 design schemes 
achieving performance 
Research on different types of 
buildings is an important 
aspect of improving and 
optimizing design indicators 

84   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
function for multi-objective 
problems. 
standards and less than 12 
modifications 
In summary, the research related to MCDM method 
has been maturely developed and widely applied in 
various industries, but in terms of building optimization 
performance, MCDM still has a large space for progress 
research, based on this, the study proposes the SB 
characteristic, introduces the sensitivity to analyze the 
variables affecting the building and improves the design 
of its design scheme. 
 
3 Construction of SMDA-based 
performance optimization model 
for sustainable building 
The MCDM algorithm, also known as the multi-criteria 
algorithm, is developed in the mid-1970s. As a decision 
analysis method, it can be useful in the phases of system 
planning, design and manufacturing, focusing on solving 
problems that may occur in the present or future. 
Conventional MCDM methods generally include two 
kinds of MCDM and multi-attribute decision making, 
which have more mature development in both theory and 
practice, and therefore the MCDM method has been more 
widely used. The study is based on the sensitivity 
multi-objective analysis to optimize the analysis of SB, 
and to explore its influence variable indicators. 
 
3.1 Construction based on the SMDA 
algorithm 
As in the real situation, architects in the construction 
industry need to coordinate the variables between 
different design objectives according to the actual needs 
in order to solve the conflict of objectives between 
multiple performances. The process is not a simple 
comparison and selection, but requires trade-offs between 
multiple objectives and related parameter adjustments. A 
SMDA is employed to investigate the design solution and 
to adjust the performance of the solution and the values 
of each parameter. The modification of the overall 
solution is carried out by comparing the values of each 
parameter's influence weight in the longitudinal direction 
of the design solution. Multi-objective performance 
design generally contains more and more complex 
attribute relationships, should be considered to use a 
higher degree of adaptation and broader sensitivity 
analysis method to optimize the BP model, considering 
the Fourier amplitude sensitivity experiments are 
relatively higher cost budget, the study uses the Sobol 
global sensitivity analysis method to calculate the 
sensitivity of the input values [13-15]. The sensitivity 
index of Sobol method is calculated as shown in Equation 
(1). 
( )
1,2,...,
1
1,2,...,
11
...
... 1
m
i ij m
j
mm
i ij m
i i j m
TS i s s s
s s s
=
=   

= + + +




+ + + =




 (1) 
In Equation (1), ( ) TS i is the total sensitivity index 
of the input variable 
i
x and 
i
s is the first order 
sensitivity index of the input variable. 
m
 is the total 
number of input variables and 
ij
s
 is the sensitivity index 
between the i th input variable and the 
j
th input 
variable. 
1,2,...,m
s
 is the sum of the first-order sensitivity 
values of the input variables. The study modifies specific 
functional relationships in accordance with the variable 
relationship between the design objectives and design 
variables, thereby enhancing the optimization of the 
functional relationship between the input and output 
values. The comprehensive performance of the design 
scheme is identified as the entry point for analysis, with 
the various factors affecting the BP design scheme 
serving as the base point. Hierarchical analysis is used to 
calculate and analyze the sensitivity matrix values of each 
design variable in terms of the performance of the design 
objectives and to calculate the weight values of the 
performance of each parameter of the original solution 
under this condition. The computational flow of the 
hierarchical analysis method is shown in Figure 1. 
 
Analyze the correlation 
between descriptive 
elements
Weight calculation based on 
feedback influence 
relationship
Constructing Extreme 
Hypermatrices
Calculate the final
 sorting result
Construct initial
 hypermatrix
Calculate
 judgment matrix
Establishment of 
judgment matrix
Start
End
 
Figure 1: The calculation process of ANP 
 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 85 
 
In Figure 1, hierarchical analysis is required to 
analyze the decision-making problem in the calculation 
process and to analyze the correlation of the relevant 
factors involved, and to judge the independence or 
correlation between the elements. In the model 
construction process, the development of a network layer 
design containing the interrelationships of elements is 
realized based on the objectives and guidelines for setting 
up a control layer. The important feature vectors of the 
quasi-measurement layer to the target layer under a single 
scenario are calculated using the hierarchical analysis 
method, and the calculation formula is shown in Equation 
(2) [16]. 
 
, ,max ,min
( ) / ( )
fi i i i s i i
c k f f f f = − −
 (2) 
In Equation (2), 
fi
c
 is the importance feature 
vector and 
i
k is the penalty function. 
i
k takes the 
value of 1 when the selected performance metrics do not 
meet the requirements of the adjustment. 
i
k takes the 
value of -0.001 when the selected performance metrics 
meet the requirements of the desired adjustment. 
i
f is 
the value assigned to the performance under the scenario, 
and 
, is
f
 is the criterion value of the i th performance 
under the individual scenario. 
,min i
f
 is the minimum 
value for the i th performance in the database. 
,max i
f
 is 
the maximum value of the i th performance in the 
database. The hierarchical analysis method is employed 
to identify the importance eigenvectors of each variable 
in the scheme layer. Subsequently, the importance 
eigenmatrix of each parameter under a specific number of 
parameters is constructed. The resulting importance 
eigenvectors of the criterion layer and the importance 
matrix of the scheme layer are calculated to determine the 
comprehensive impact weights of all the variables on the 
original scheme. This process yields Equation (3). 
1 1 1
2 2 2
12
12
12
12
...
...
[ , ,..., ]
n n n
f f f
n
f f f
n
f f fn
f f f
n
c c c
c c c
w c c c
c c c



=




 (3) 
In Equation (3), 
w
 is the comprehensive impact 
weight of all variables on the original program, 
1 1 1
2 2 2
12
12
12
...
...
n n n
f f f
n
f f f
n
f f f
n
c c c
c c c
c c c







 is the importance matrix under the 
program level, and the rest of the explanation is 
consistent with the meaning of the previous section. For 
the degree of contribution of the first design variable to 
the current target program, the formula is shown in 
Equation (4). 
12
12
* * ... *
m
m
f ff
j f j f j f j
w c c c c c c = + + + (4) 
In Equation (4), 
j
w
 is the contribution value of the 
design variables to the degree of comprehensive 
performance improvement of the current target program, 
and the rest of the parameters are interpreted in the same 
way as before. According to the size of the obtained 
contribution value can further determine the improvement 
ability and enhancement effect on the comprehensive 
performance of the current research program when the 
reference variable changes. If the 
j
w
 value is >0, the 
value of the variable should be reduced accordingly, and 
vice versa, the value of the variable should be increased 
accordingly. 
 
3.2 Construction of SMDA-based 
performance optimization algorithm for 
sustainable building 
After constructing the completed SMDA algorithm, the 
study has been able to calculate the weighting effects of 
the parameter factors in the original scheme on the 
scheme, but this does not directly provide the architect 
with a certain performance-attainment design solution. 
Therefore, the study weighs each objective on the basis of 
multi-objective optimization problem decision making, 
fuses the SMGA algorithm with genetic algorithm (GA), 
and proposes a method of sensitivity multi objective 
decision fusion genetic algorithm (SMGA) an intelligent 
optimization algorithm that will use the objective 
expectations instead of subjective weights weighting to 
the computation of the fitness function. Since the GA 
fused with SMGA algorithm is capable of solving the 
involved problems at runtime by imitating the relevant 
principles and phenomena and using heuristics to 
optimize the problem without prior knowledge of the 
mathematical characteristics of the solution of the 
problem to be optimized, its high robustness and 
practicability have led to a wide range of applications of 
the GA in the design of building solutions, operation and 
maintenance of equipment and energy. The functional 
expression defining the multi-objective adaptation 
function ( ) Gx is shown in Equation (5). 
 ( )
1
max *
m
ii
i
G x k g
=
=

 (5) 
In Equation (5), 
i
g is the normalized fitness value 
of the i th performance objective, and 
i
k is the penalty 
function of each performance objective. When i fails, 
i
k takes the value of 10000, and vice versa 
i
k takes the 
value of 1. At a later stage, the normalized fitness values 
i
g of all the performance objectives are summed to 
obtain the fitness function of the multi-objective problem. 
At the same time, the operator is selected for the 
unqualified performance i and the sifting out of the 
solutions of the infeasible solution scheme is carried out. 
In the GA variation session, for the design scheme 
12
[ , ,..., ]
n
x x x x = constructed by the 
n
 dimensional 

86   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
decision variables, the weight vector 
12
[ , ,..., ]
n
w w w w = 
of the influence of each variable on the current scheme is 
calculated using the SDMA algorithm previously used in 
the study, by changing the evolution. The maximum 
variable of lucidity in the process, which in turn improves 
the objective function. Roulette is performed on the 
merged population 
R
 thus determining the operator, 
selecting the number of individuals 
p
N
 to perform 
variation and crossover operations on R , calculating the 
fitness function 
()
j
Fx
 and normalizing it as shown in 
Equation (6). 
 
min
max min
()
'( )
j
j
F x f
Fx
ff
−
=
−
 (6) 
In Equation (6), 
'( )
j
Fx
 is the normalized fitness 
function of the interpretation scheme and 
()
j
Fx
 is the 
fitness function. 
min
f is the performance minimum of 
the explained solution and 
max
f is the performance 
maximum of the explained solution. After obtaining the 
normalized fitness function of the solution, the selection 
probability of each solution is further calculated and its 
functional expression is shown in Equation (7). 
 2
1
'( )
()
'( )
j
j Np
j
j
Fx
px
Fx
=
=

 (7) 
In Equation (7), 
p
N
 is the number of selected 
individuals. Using the same idea to solve the cumulative 
probability, the functional expression is shown in 
Equation (8). 
 
1
( ) ( )
j
jr
r
q x p x
=
=

 (8) 
In Equation (8), r is a member in population R 
and ()
r
px is the selection probability of the selected 
member r . Based on the 
p
N
 explanatory schemes 
obtained from the selection operator, further mutation 
operations are carried out, and in order to prevent the 
SMGA proposed by the study from over-controlling the 
individual's best solution scheme and thus falling into a 
local optimal solution situation, the study stratifies the 
p
N
 explanatory schemes obtained. This is formulated as 
follows: 2/3 of the 
p
N
 explanation scheme is used for 
the previously constructed SMDA algorithm for the 
mutation operation, and the remaining 1/3 
p
N
 
explanation scheme is used for the regular mutation 
operation. In the variation operation of SMDA, for the 
j
th solution of the scheme 
12
[ , ,..., ]
n
j j j j
x x x x = , the 
weight 
12
[ , ,..., ]
n
j j j j
w w w w = is calculated according to 
Equation (4) mentioned earlier in the study, which in turn 
calculates the normalized influence weight of the k th 
design variable 
k
j
x . The calculation formula is shown in 
Equation (9). 
 
min max min
' ( ) / ( )
kk
jj
w w w w w = − −
 (9) 
In Equation (9), '
k
j
w is the normalized impact 
weight of the design variables and 
k
j
w is the weight of 
the design variables before variance crossover. 
max
w is 
the maximum value of the impact weights of all design 
variables and 
min
w is the minimum value of the impact 
weights of all design variables. A random number d is 
generated for comparison for any design variable 
k
j
x in 
the closed interval from 0 to 1. If the randomly generated 
value d is less than or equal to the normalized weights, 
the associated k th actual variable is variant. When the 
normalized weights are greater than 0, the corresponding 
design variables should be reduced in value accordingly 
and a new design variable is randomly generated at the 
min
[ , ]
k
j
dx as a new design variable participating in the 
mutation crossover. The conventional variation operation 
is to use a fixed rate of variation instead of normalized 
weights when performing variation operation on a single 
design solution [17]. When the value of the selected 
random number of values in the selected interval is less 
than the fixed rate of variation, the same randomly 
generates a new design variable in 
min max
[ , ] kk for the 
variation operation [18, 19]. The flow of the SMGA 
algorithm proposed in the study is shown in Figure 2. 
 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 87 
 
Start
Merge previous and 
subsequent generations of 
populations
Applying selection operators 
to the merged population
SADM calculates the rate of 
variation for each variable in 
each scheme
Calculate the objective function 
fx and fitness function F (x) of 
the population
Population fitness ranking, 
adding the solution with the 
highest fitness to the result 
library
End
Evolution count t=0, initialize 
population
Mutation operation to obtain 
P t ’ ，Cross intersection 
operation, obtaining P t "
Generate a new generation of 
population P t+1 =P t "
Is the current generation
 number> the specified 
generation number?
Delete schemes that do 
not meet constraints and 
duplicate schemes in D
Output D
Y
Y
t
=
t
+
1
 
Figure 2: Schematic diagram of SMGA algorithm flow 
 
In Figure 2, after the mutation operation and part of 
the regular operation using the SDMA algorithm, it is 
necessary to carry out the corresponding population 
crossover operation on the population after the end of the 
mutation operation. For the 
j
th solution 
12
[ , ,..., ]
n
j j j j
x x x x = obtained using the SDMA algorithm 
a number 
c
 is randomly generated in the interval [0,1]. 
if the selected value 
c
 is smaller than its corresponding 
selection probability 
c
p , a crossover operation needs to 
be carried out on it, and the parent generated by the 
crossover operation for the 
1
c th solution is 
12
1 1 1 1
[ , ,..., ]
n
c c c c
x x x x = . subsequently, an indexed value 
2
c 
is randomly generated in 
[1, ]
p
N
 and another crossover 
parent is generated 
12
2 2 2 2
[ , ,..., ]
n
c c c c
x x x x = , and a 
randomly generated bit value 
c
v in [1,n] is used to 
confirm the position of the selected parent scheme at 
crossover. In turn, a new scheme 
1
'
c
x is obtained, 
whose functional expression is shown in Equation (10). 
1 2 1
1 1 1 1 2 2
' [ , ,..., , ,..., ]
vc vc n
c c c c c c
x x x x x x
+
= (10) 
Perform crossover operation for each solution in the 
population to get a new generation of population. Set the 
number of iterations to loop from computing the 
objective function and fitness function to the crossover 
operation. Determine whether the number of iterations 
has reached the maximum number of iterations, and thus 
determine whether to stop the optimization to output the 
optimal solution set. Finally output the optimal solution 
set and process the result database to remove the 
solutions and performances in the database that do not 
meet the set expectation and output the optimal solution 
set of the remaining solutions. 
4 Examination of the effectiveness of 
the application of sustainable 
building performance optimization 
design 
In the method design section, the research first constructs 
a multi-objective decision-making algorithm based on 
sensitivity, and calculates the matrix values and weight 
values of the target performance under the variables in 
the architectural scheme design. Concurrently, research is 
conducted on the balancing of various objectives based 
on multi-objective optimization problem decision-making, 
integrating the SMGA algorithm with the GA, and using 
an intelligent optimization algorithm to adjust weight 
values, thereby avoiding the influence of subjective 
settings on the results. To validate the efficacy of the 
proposed algorithm, a case study is conducted on a 
building in a cold climate to optimize its energy 
consumption, lighting, and other multi-objective issues. 
The requisite parameters for the algorithm are then 
defined. The study set the population size and the number 
of iterations to be 50, the variation rate to be 0.1, and the 
crossover rate to be 0.9. The algorithm is optimized for 
the population on the basis of the initial scheme in the 
same experimental environment, and the number of 
experiments is 20. During the experiment, the 
performance metrics of the GA before and after the 
improvement are collected and the results are shown in 
Table 2. 
 
 
 
Table 2: Performance results of GA before and after optimization 

88   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
Number of 
experiments 
Convergence 
speed 
Optimal solution 
size 
The diversity of decision 
space 
Effectiveness 
1 11 4 8 14 0.057 0.071 8.874 9.053 
2 12 7 7 7 0.057 0.055 4.516 7.535 
3 22 6 5 9 0.06 0.073 7.144 9.748 
4 12 6 5 11 0.05 0.054 3.962 8.741 
5 7 2 6 12 0.066 0.063 4.823 9.721 
6 13 9 6 11 0.062 0.061 7.203 10.208 
7 17 6 9 9 0.065 0.065 8.203 9.57 
8 16 6 5 8 0.06 0.055 7.027 8.128 
9 6 2 7 10 0.054 0.066 7.701 10.174 
10 2 7 10 5 0.061 0.049 5.627 7.475 
11 15 2 4 8 0.04 0.068 5.8 9.102 
12 10 3 8 11 0.057 0.071 8.086 9.065 
13 34 4 4 10 0.064 0.068 2.678 10.397 
14 12 4 6 8 0.068 0.042 10.292 9.057 
15 2 6 6 12 0.059 0.066 6.451 8.861 
16 37 6 4 12 0.052 0.077 3.869 10.351 
17 23 3 7 9 0.064 0.067 7.027 8.633 
18 18 8 5 11 0.059 0.054 7.397 9.238 
19 11 5 5 12 0.065 0.069 4.62 7.123 
20 11 4 8 14 0.057 0.071 8.874 9.053 
Average value 14 4.8 5.85 9.45 0.056 0.059 6.065 8.609 
 
In Table 2, the convergence speed of SAGA 
algorithm is significantly better than that of GA, and its 
average value reaches 4.8, which indicates that it needs 
an average of 4.8 iterations to find the attainment result in 
the program to be completed, while GA needs 14 
iterations. In addition, in the optimal solution size and 
decision space diversity, the average value of SAGA 
algorithm reaches 9.45 and 0.059, which is significantly 
higher than that of GA's 5.85 and 0.056, which indicates 
that the improved SAGA algorithm can effectively avoid 
falling into the local optimum, and its probability of 
finding the optimal solution in the global is increased.  
 
 
The SAGA algorithm can obtain the performance of the 
average number of design solutions up to the standard is 
5.58, and in the analysis of the effectiveness of the 
algorithm, the average value of the SAGA algorithm is 
8.609 is much larger than the GA's 0.065, which shows a 
better performance advantage. A comparison is to be 
made between the proposed optimization SAGA 
algorithm and the decision-making method. Two 
comparison scenarios are to be set for different 
orientations of the building. A comparison of the 
performance results of different algorithms in building 
scheme design is presented in Figure 3. 
 
200
250
300
350
400
450
500
550
600
1 2 3 4 5 6 7 8 9 10
200
250
300
350
400
450
500
550
600
1 2 3 4 5 6 7 8 9 10
Total energy consumption/GJ Total energy consumption/GJ
Solving time/h
Solving time/h
Reference [10]
Reference [11]
Research method
Reference [10]
Reference [11]
Research method
 
Figure 3: Pareto optimal solution results for different algorithms 
 
The results presented in Figure 3 demonstrate that 
the convergence of the Pareto optimal frontier solution 
demonstrated by the SAGA algorithm in southbound 
buildings is significantly higher than that observed in 
reference [10]. The discrepancy between the two 
solutions is more pronounced when the time-energy 
consumption increases. Furthermore, the solution set 
distribution in reference [10] is more dispersed, 
exhibiting poor convergence. The SAGA algorithm 
reduces the solving time by 13.25% and 5.39% compared 
to references [10] and [11], respectively, while exhibiting 
a relatively minor increase in energy consumption. In the 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 89 
 
architectural design of the Northern Dynasties, the 
convergence of the SAGA algorithm is demonstrably 
higher than that of reference [10], and the discrepancy 
with reference [11] is relatively minor. This discrepancy 
may be attributed to the fact that reference [11] is capable 
of considering building parameters to a certain extent, yet 
its resource consumption during calculation is relatively 
high. In conclusion, the SAGA algorithm proposed in the 
study demonstrates satisfactory convergence performance. 
The decision diversity of the SAGA algorithm is further 
analyzed to obtain the variance of different decision 
variables, and the results are shown in Figure 4. 
0.07
0.06
0.05
0.04
Variance
B C A D E F G H I J K L M N O
0.03
0.02
0.01
0
Variable
GA 
SAGA 
Aspect ratio
Facing
Floor height
North facing window to wall ratio
West facing window to wall ratio
South facing window to wall ratio
East facing window to wall ratio
North facing window wall height
West facing window wall height
South facing window wall height
A
B
C
D
E
F
G
H
I
J
East facing window wall height
South facing sun shading angle
Sunshade construction angle
Sunshade construction materials
Glass heat transfer coefficient
K
L
M
N
O
 
Figure 4: The variance results of two algorithms on decision variables 
 
In Figure 4, overall, it seems that the SAGA 
algorithm has less variance results than the GA for the 
building target variables, in which the SAGA algorithm is 
more sensitive to the sill height, shading angle, shading 
member angle and material, and glass heat transfer 
coefficient (GHTC)) variables, and the diversity in the 
optimal solution is lower. Improving GA can effectively 
increase the sensitivity of the building variables. The 
study first analyzes the extraction effect on the acquired 
sample data and examines the variable extraction effect 
of the improved algorithm proposed in the study, and the 
results are shown in Figure 5. 
 
0 32 64 96 128
0.8
1.0
1.2
1.4
0.0
0.5
1.0
1.5
2.0
0 16 32 48 64
Sampling point
Precision 
Precision 
Sampling point
(a) Before improvement (b) After improvement
2.5
 
Figure 5: Variable extraction effect of improved algorithm 
 
In Figure 5, the improved algorithm shows a better 
sample accuracy effect, and its overall node distribution 
is more uniform, which can better identify the data 
features. The study takes the standard floor building in 
the cold northern region as an example, and optimizes its 
energy consumption and lighting with the help of Energy 
Plus software and Radiance software. Application results 
supplement: The proposed method will be used to 
analyze the error rate of variable data for buildings of 
different heights, as shown in the Figure 6. 
 

90   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
0 5 10 15 20 25
10
-4
10
-2
10
0
Number of experiments
Value
Train
Validation
Test
(a) Middle and low rise buildings
10
-2
10
-1
10
0
0 5 10 15 20 25
Number of experiments
Value
(b) High-rise buildings
Train
Validation
Test
 
Figure 6: Error rate of variable data for buildings with different floor heights 
 
The results presented in Figure 6 indicate that in 
mid- to low-rise buildings, the error rate value of the 
analysis of building parameter variables in the study 
shows a decreasing trend. Furthermore, there is a small 
error amplitude between the analysis and the actual 
effective value, with the minimum error value less than 
10-4. In high-rise buildings, the algorithm proposed in the 
study is capable of obtaining building variable data with a 
minimum error value of less than 10-2. The research is 
based on the building model construction platform, in 
which the parameter programming platform is 
Grasshopper, with the help of Honeybee+ and other 
plug-ins for performance simulation and analysis, and the 
rest of the programming language is Python. The initial 
variable design of the building model is shown in Table 
3. 
 
 
Table 3: Variables of building models 
Variable Initial plan First improvement Second improvement 
A 1 1 1 
B 0 0 0 
C 4.5 4.5 3.7 
D 0.48 0.48 0.48 
E 0.36 0.36 0.36 
F 0.48 0.48 0.48 
G 0.36 0.36 0.36 
H 1 1 1 
I 1 1 1 
J 1 1 1 
L 1 1 1 
M 0 0 0 
N Aluminum alloy Aluminum alloy Aluminum alloy 
O 2.4 1.2 2.4 
 
On the initial building scheme, its storey height and 
GHTC are optimized and designed respectively. In the 
first improvement, the GHTC is adjusted downward from 
2.4 to 1.2, and in the second improvement, the storey  
 
height is varied from 4.5 to 3.7. The results of the 
performance of the scheme under the improvement are 
obtained and shown in Figure 7. 
 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 91 
 
91.0
90.5
91.5
92.0
Energy consumption (KWh/m
2
)
93.0 80
70
60
50
40
30
4
5
6
7
8
9
10
11
0 1 2 3
0 1 2 3
0 1 2 3
(a) Energy 
Number of experiments
92.5
sDA(%)
ASE(%)
(b) sDA (c) ASE
Not meeting the 
standard value
Compliance value
Standard value
Number of experiments Number of experiments
Not meeting the 
standard value
Compliance value
Standard value
Not meeting the 
standard value
Compliance value
Standard value
 
Figure 7: Performance results of the improved scheme 
 
In Figure 7, after the improvement of the initial 
scheme, the energy consumption of the building is 
significantly reduced, and its value is reduced to reach 
89.89 KWh/m2 and its performance is up to standard. 
Among them, the sDA indicator has a small decrease 
after the program modification, but it is still in the range 
of meeting the standard. The above results show that the 
modified program is effective. Subsequently, the variable 
impact weights of the program are analyzed and the 
results are shown in Table 4. 
 
 
 
 
Table 4: Variable influence weights of improvement plans 
Variable Initial plan First improvement Second improvement 
A -0.00689 -0.002564 -0.03809 
B 0.000064 0.00439 -0.031136 
C 0.152747 0.157073 0.121547 
D 0.050189 0.054515 0.078989 
E 0.046116 0.050442 0.014916 
F 0.0923524 0.0966784 0.0611524 
G 0.132656 0.136982 0.101456 
H -0.000253 0.004073 -0.031453 
I -0.000286 0.00404 -0.031486 
J -0.000006 0.00432 -0.031206 
L -0.000031 0.004295 -0.031231 
M 0.002013 0.006339 -0.029187 
N -0.008959 -0.004633 -0.040159 
O 0.227903 0.272229 0.296703 
 
In Table 4, the values of GHTC for the scheme 
design with the help of SADA algorithm have weighted 
values of 0.272229W/m2K and 0.296703W/m2K for the 
first improvement and the second improvement, 
respectively. The solar coefficient of the glass has a more  
 
 
obvious impact in the optimization of the scheme design, 
so the influence of this indicator should be paid attention 
to in the future research. The predicted results of the sDA 
indicator are analyzed and the results are shown in Figure 
8. 
 

92   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
Normalized value
0 0.2 0.4 0.6 0.8 1.0
1.0
0.5
0
Normalized value
 
Figure 8: Fitting results 
 
In Figure 8, the sDA indicator is analyzed in the 
prediction, which has a high consistency with the actual 
value, and its value is basically distributed on both sides 
of the curve. The proxy results of ASE indicators are 
further analyzed, and their results are shown in Table 5. 
 
Table 5: Proxy of ASE indicators 
Variable 1 2 3 4 5 
R2 
0.689 0.668 0.637 0.648 0.627 
0.605 0.584 0.553 0.564 0.543 
RMSE 
0.076 0.055 0.024 0.035 0.014 
0.81 0.789 0.758 0.769 0.748 
MAE 
0.059 0.038 0.007 0.018 -0.003 
0.064 0.043 0.012 0.023 0.002 
 
In Table 5, the error results of ASE results under 
different number of experiments are small and the fitted 
values are less than 0.70 in both training and test sets, and 
the mean values of RMSE and MAE values reach 0.078 
and 0.063 in the training set, and 0.061 and 0.067 in the 
test set. A total of 30 architectural professionals and 
engineers are also invited to modify the initial under test  
 
The design solutions are modified, and their modification 
criteria are based on their subjective experience in 
obtaining the rest of the metrics, except for the three 
properties mentioned in the study. The modification idea 
of the comparison is to modify the design variables by the 
system modification suggestions generated by the SADM 
algorithm, and the data statistics of the modifications 
under the two approaches are shown in Figure 9. 
 
45
40
35
30
25
20
15
10
5
0
5 0 10 15 20 25 30
Number of modifications
Tester ID
 After SADM
 Before SADM
 
Figure 9: Number of modifications with and without the guidance of SADM method 
 
In Figure 9, when the testers optimize the scheme 
without system recommendations, the number of 
modifications are more than 15 times, of which the 
maximum value is 43 times, and the average number of 
modifications is 22 times. While in the program 
recommended optimization suggestions, the tester's 
modification times are significantly reduced, the overall 
curve of the large fluctuations are significantly reduced, 
the number of modifications are less than 12 times, the 
minimum number of modifications is reached 3 times. 
Subsequently, the energy consumption and lighting test 
results of the proposed algorithm are analyzed. Three 
cities are selected for analysis: Beijing, Harbin, and 
Changchun. The results are shown in the Figure 10. 
 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 93 
 
Test value Training value
6
0
2
4
6
0
2
4
Mean square error
Mean square error
Value
Training value
Value
Test value
Beijing
Changchun
Harbin
Beijing
Changchun
Harbin
(a)Energy consumption (b)Daylighting
 
Figure 10: Test results of mean square error of SAGA algorithm in typical cold regions 
 
Figure 10(a) depicts the mean square error of the 
energy consumption prediction results. It can be observed 
that the SAGA algorithm exhibits high accuracy in 
testing building energy consumption in three cities, with 
the error value between the test results and actual results 
tending to be close to 0. In the lighting results (Figure 
10(b)), the SAGA algorithm also demonstrated favorable 
test results, with a maximum mean square error of 4 and a 
minimum value of 1.6. Specific analysis of the energy 
consumption of the building after optimization, the 
results are shown in Figure 11. 
 
-14
-12
-10
-8
-6
-4
-4
-2
0
2
4
8
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Time(weeks)
Temperature difference inside and 
outside the building (℃)
(a) Before improvement
50
45
500 1000 1500 2000 2500
40
35
30
25
20
15
10
5
0
225 0 725 1225 1725 2250 2750
Time(h)
Energy consumption(KW/h)
Energy consumption(KW/h)
(b) After improvement
 
Figure 11: Water temperature change and energy consumption in hydronics 
 
When the outdoor temperature fluctuates greatly 
with obvious nodal ups and downs, the temperature 
difference between the inside and outside of the building 
decreases, resulting in lower energy consumption. This 
also leads to a smaller overall volatility of the curve, 
effectively ensuring the building's low energy 
consumption. 
5 Discussion 
This research examines the optimization of SB 
performance design based on sensitivity multi-objective 
decision-making and its subsequent application analysis. 
The results presented in Table 2 and Fig. 3 demonstrated 
that the proposed SAGA algorithm exhibits superior 

94   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
convergence compared to traditional GA and other 
comparative algorithms. The SAGA algorithm required 
only 4.8 iterations to reach the optimal value, whereas the 
GA requires more than 10 iterations. The GA exhibits a 
tendency to converge poorly due to its proclivity to fall 
into local optima. In response to the limitations of the GA, 
scholars such as Saryazdi SME have also combined the 
GA with artificial neural networks to achieve 
construction design optimization [20]. References [10] 
and [11] employed multi-objective decision-making 
methods based on decision-maker preferences and 
building information model fusion for the evaluation of 
BP. Nevertheless, their reliance on subjective expert 
decision-making and high resource consumption resulted 
in suboptimal overall convergence of the two references. 
The Pareto solution set distribution in reference [10] was 
relatively dispersed, and the SAGA algorithm reduced the 
solution time compared to references [10] and [11] by 
13.25% and 5.39%, respectively. In the context of data 
processing, the SAGA algorithm demonstrated superior 
accuracy in sample extraction when the variance of the 
building target variable was less than that of the GA. 
Moreover, the variable data was smaller in buildings of 
different heights, and the minimum error values were less 
than 10-4 and 10-2 in mid-to-low-rise and high-rise 
buildings, respectively. This outcome can be attributed to 
the fact that a sensitivity-based analysis of building 
parameters allows for a more comprehensive 
consideration of factors such as lighting, floor height, and 
building materials, which in turn ensures a high degree of 
consistency between test results and actual outcomes. 
This result was analogous to that presented in literature 
[13], which demonstrated the efficacy of multi-objective 
decision analysis on variables in improving the 
effectiveness of design outcomes. Following the 
implementation of an enhanced initial plan, the energy 
consumption of the building was reduced to 89.89 
kWh/m2, thereby meeting the established performance 
standards. The predictive analysis of sDA indicators 
demonstrated a high degree of consistency with the actual 
values. The results of the building energy consumption 
analysis demonstrated that the optimized plan effectively 
ensures low energy consumption of the building. This 
result was consistent with the findings of previous 
literature. Among these, literature [21] employed the 
simulated annealing algorithm and sensitivity analysis to 
address multi-objective issues in architectural design, 
while literature [22] utilized an enhanced Garson 
algorithm and sensitivity based on parameter design to 
quantify building envelope structures. Both studies have 
demonstrated the efficacy of sensitivity analysis in the 
context of building parameters. 
6 Conclusion 
The study introduces SMDA to analyze the performance 
optimization of SB and examine the effectiveness of its 
application. The results indicated that the convergence 
speed of SAGA algorithm was significantly better than 
that of GA, and its average value reached 4.8, which 
indicated that it needed an average of 4.8 iterations to be 
completed to find the attainment result in the program. 
Furthermore, in terms of optimal solution size and 
decision space diversity, the average value of the SAGA 
algorithm reached 9.45 and 0.059, which is significantly 
higher than that of the GA, which is 5.85 and 0.056. The 
SAGA algorithm was more sensitive to sill height, 
shading angle, shading member angle, and material, as 
well as GHTC variables. In the case study, after the 
improvement of the initial scheme, the energy 
consumption of the building was significantly reduced, 
and its value reduction reached 89.89 KWh/m2. The 
values of the building's GHTC were weighted with values 
of 0.272229W/m2K and 0.296703W/m2K for the first 
and second improvements, respectively. The predicted 
results of the sDA metrics exhibited a high degree of 
consistency with the actual values, the fitted values of the 
ASE were less than 0.70 in both the training and test sets, 
and the mean values of the RMSE and MAE values 
reached 0.078 and 0.063 in the training set and 0.061 and 
0.067 in the test set. The efficiency of the testers in 
carrying out the programmed multi-objective 
optimization was significantly improved with the 
suggested modifications of the algorithm. The proposed 
method of the study can effectively improve the 
performance of optimal building design and its 
performance for multi-objective optimization problems is 
better. Enriching the types of buildings studied and 
further supplementing the variable indicators are 
important elements for the study to focus on in the future. 
References
 
[1] A. Mota, P. Vila, R. Albuquerque, L. Costa, and J. 
Bastos, “A framework for time-cost-quality 
optimization in project management problems using 
an exploratory grid concept in the multi-objective 
simulated-annealing,” International Journal of 
Information Technology and Decision Making, vol. 
10, pp. 320-327, 2021. 
https://doi.org/10.1142/S0219622021500322 
[2] B. Xu, Y. Sun, X. Huang, P. Zhong, F. Zhu, J. Zhang, 
X. Wang, G. Wang, Y. Ma, Q. Liu, H. Wang, and L. 
Guo, “Scenario-based multiobjective robust 
optimization and decision-making framework for 
optimal operation of a cascade hydropower system 
under multiple uncertainties,” Water Resources 
Research, vol. 54, no. 8, pp. 5763-5784, 2022. 
https://doi.org/10.1029/2021WR030965 
[3] B. Lin, H. Chen, Q. Yu, X. Zhou, S. Lu, Q. He, and Z. 
Li, “MOOSAS - a systematic solution for multiple 
objective building performance optimization in the 
early design stage,” Building and Environment, vol. 
200, pp. 10-16, 2021. 
https://doi.org/10.1016/j.buildenv.2021.107929 
[4] S. R. Nabavi, Z. Wang, and G. P. Rangaiah, 

Construction of Sustainable Building Performance Optimization… Informatica 48 (2024) 81–96 95 
 
“Sensitivity analysis of multi-criteria 
decision-making methods for engineering 
applications,” Industrial and Engineering Chemistry 
Research, vol. 62, no. 17, pp. 6707-6722, 2023. 
https://doi.org/10.1021/acs.iecr.2c04270 
[5] R. J. I. Basten, M. C. Van der Heijden, and J. M. J. 
Schutten, “A minimum cost flow model for level of 
repair analysis,” International Journal of Production 
Economics, vol. 133, pp. 233-242, 2011. 
https://doi.org/10.1016/j.ijpe.2010.03.025 
[6] B. Sarkar, and A. Biswas, “A multi-criteria decision 
making approach for strategy formulation using 
Pythagorean fuzzy logic,” Expert Systems: The 
International Journal of Knowledge Engineering, 
vol. 39, no. 1, pp. 812-825, 2022. 
https://doi.org/10.1111/exsy.12802 
[7] P. Zurcher, H. Shirahata, S. Badr, and H. Sugiyama, 
“Multi-stage and multi-objective decision-support 
tool for biopharmaceutical drug product 
manufacturing: Equipment technology evaluation,” 
Chemical Engineering Research and Design: 
Transactions of the Institution of Chemical 
Engineers, vol. 161, pp. 240-252, 2020. 
https://doi.org/10.1016/j.cherd.2020.07.004 
[8] D. Patil, and R. Gupta, “GIS-based multi-criteria 
decision-making for ranking potential sites for 
centralized rainwater harvesting,” Asian Journal of 
Civil Engineering: Building and Housing, vol. 24, 
no. 2, pp. 497-506, 2023. 
https://doi.org/10.1007/s42107-022-00514-z 
[9] B. Lin, H. Chen, Y. Liu, Q. He, and Z. Li, “A 
preference-based multi-objective building 
performance optimization method for early design 
stage,” Building Simulation (English), vol. 014, no. 
003, pp. 477-494, 2021. 
https://doi.org/10.1007/s12273-020-0673-7 
[10] D. Utkucu, and H. Sozer, “Building performance 
optimization throughout the design-decision process 
with a holistic approach,” Journal of Architectural 
Engineering, vol. 29, no. 1, pp. 22-34, 2023. 
https://doi.org/10.1061/(ASCE)AE.1943-5568.0000
573 
[11] C. Xie, and D. Wang, “Multi-objective 
cross-sectional shape and size optimization of S-rail 
using hybrid multi-criteria decision-making 
method,” Structural and Multidisciplinary 
Optimization, vol. 62, no. 6, pp. 3477-3492, 2020. 
https://doi.org/10.1007/s00158-020-02651-y 
[12]S. M. Jafari, M. R. Nikoo, O. Bozorg-Haddad, N. 
Alamdari. R. Farmani, and A. H. Gandomi, “A 
robust clustering-based multi-objective model for 
optimal instruction of pipes replacement in urban 
WDN based on machine learning approaches,” 
Urban Water Journal, vol. 20, no. 5/6, pp. 689-706, 
2023. 
https://doi.org/10.1080/1573062X.2023.2209063 
[13] Z. Han, X. Li, J. Sun, M. Wang, and G. Liu, “An 
interactive multi-criteria decision-making method 
forbuilding performance design,” Energy and 
Buildings, vol. 282, no. 3, pp. 14-17, 2023. 
https://doi.org/10.1016/j.enbuild.2023.112793 
[14] C. Zong, M. Margesin, J. Staudt, F. Deghim, and W. 
Lang, “Decision-making under uncertainty in the 
early phase of building facade design based on 
multi-objective stochastic optimization,” Building 
and Environment, vol. 226, no. 12, pp. 9-14, 2022. 
https://doi.org/10.1016/j.buildenv.2022.109729 
[15] D. Panda, S. Dehuri, and A. K. Jagadev, 
“Multi-objective artificial bee colony algorithms and 
chaotic-TOPSIS method for solving flowshop 
scheduling problem and decision making,” 
Informatica, vol. 44, no. 2, pp. 241-262. 
https://doi.org/10.31449/inf.v44i2.2616 
[16] D. H. Tran, “Optimizing time-cost in generalized 
construction projects using multiple-objective social 
group optimization and multi-criteria 
decision-making methods,” Engineering 
Construction and Architectural Management, vol. 27, 
no. 9, pp. 2287-2313, 2020. 
https://doi.org/10.1108/ECAM-08-2019-0412 
[17] A. Ebrahimi, S. Attar, and F. B. Moghadam, “A 
multi-objective decision model for residential 
building energy optimization based on hybrid 
renewable energy systems,” International Journal of 
Green Energy, vol. 18, no. 6/10, pp. 775-792, 2021. 
https://doi.org/10.1080/15435075.2021.1880911 
[18] X. Guo, J. Cao, and D. Ren, “Multi-objective 
decision making of natural gas distribution 
optimization considering clean heating-evidence 
from Beijing,” Energy Sources, Part A. Recovery, 
Utilization, and Environmental Effects, vol. 45, no. 
1, pp. 2253-2266, 2023. 
https://doi.org/10.1080/15567036.2023.2185701 
[19] H. Chandravva, and M. Hr, “Comprehensive dataset 
building and recognition of isolated handwritten 
Kannada characters using machine learning 
models,” Artificial Intelligence and Applications, 
vol. 1, no. 3, pp. 179-190, 2023. 
https://doi.org/10.47852/bonviewAIA3202624 
[20] S. M. E. Saryazdi, A. Etemad, A. Shafaat, and A. M. 
Bahman, “Data-driven performance analysis of a 
residential building applying artificial neural 
network (ANN) and multi-objective genetic 
algorithm (GA),” Building and Environment, vol. 
225, pp. 109633, 2022. 
https://doi.org/10.1016/j.buildenv.2022.109633 
[21] N. R. M. Sakiyama, J .C. Carlo, L. Mazzaferro, and 
H. Garrecht, “Building optimization through a 
parametric design platform: using sensitivity 
analysis to improve a radial-based algorithm 
performance,” Sustainability, vol. 13, no. 10, pp. 
5739, 2021. https://doi.org/10.3390/su13105739 
[22] P. Ni, W. Wang, H. Zheng, and W. Ji, “Sensitivity 
analysis of design parameters of envelope enclosure 
performance in the dry-hot and dry-cold areas,” 
Applied Mathematics and Nonlinear Sciences, vol. 8, 

96   Informatica 48 (2024) 81–96                                                                 Q. Fang et al. 
 
no. 1, pp. 1-13, 2021. 
https://doi.org/10.2478/amns.2021.2.00033