https://doi.org/10.31449/inf.v47i10.5407 Informatica 47 (2023) 103 –108 103 
Study on the Effect of Decompression Music on Anxiety Patients 
Based on Emotion Classification Algorithm 
Dan Zhang 
Art College, Shandong University of Aeronautics, Binzhou, Shandong 256600, China 
E-mail: z3d5hf@yeah.net 
 
Keywords: emotion classification algorithm, music therapy, anxiety, efficacy 
 
Received: November 8, 2023 
The relief of anxiety is important for physical health. This paper briefly introduces music therapy and 
the multimodal fusion-based emotion classification algorithm used to accurately select appropriate 
emotion music for music therapy. Afterwards, the algorithm was first tested in an experiment, and then 
the efficacy of music therapy was tested with 80 contemporaneous anxiety patients from XX Hospital in 
Binzhou City, Shandong Province. The results showed that the multimodal fusion-based music emotion 
classification algorithm was sufficiently accurate; conventional medication and care were also effective 
in relieving the anxiety symptoms, but music therapy further alleviated patients’ anxiety. 
Povzetek: Opisana sta glasbena terapija in algoritem za klasifikacijo čustev, oboje testirano na 
anksioznih pacientih. 
 
1 Introduction 
In today's fast-paced life, anxiety and stress have become 
common problems faced by many people. If these 
problems are not addressed or alleviated in a timely 
manner, over time, the negative effects caused by 
psychological problems will gradually affect 
physiological state [1]. Generally, anxiety and stress can 
be relieved by self-regulation of one ’s psychological state 
or distraction by other things without serious effects, but 
when anxiety cannot be relieved autonomously, it may 
develop into a mental illness. While medication can treat 
anxiety [2], it only relieves physical symptoms and has 
side effects. In addition to medication, music can also be 
used as a non-pharmacological treatment in the process of 
anxiety treatment. Music uses its own specific rhythm to 
help anxiety patients relax and relieve anxiety in the 
treatment process [3]. By using the music emotion 
classification algorithm, an intelligent algorithm, 
therapists can more accurately select emotionally 
appropriate music to achieve better treatment outcomes. 
This paper briefly introduced music therapy and the 
multimodal fusion-based emotion classification algorithm 
used to accurately select emotionally appropriate music in 
music therapy. The multimodal fusion-based emotion 
classification algorithm was first tested in an experiment, 
followed by a test of the efficacy of music therapy with 80 
contemporaneous anxiety patients from XX Hospital in 
Binzhou City, Shandong Province. In previous studies, 
music therapy has been examined for its therapeutic 
effects on mental disorders. This article takes a 
perspective of selecting appropriate therapeutic music and 
utilizes an emotion classification algorithm to choose 
suitable music for treating anxiety. By employing this 
algorithm, it aims to enhance the treatment effectiveness 
by selecting music that is better suited for patients. 
2 Related works 
Authors Main content Results 
Lu et al. They investigated the 
efficacy of music 
therapy on sleep 
disorders in patients 
with schizophrenia. 
The music 
therapy was 
effective in 
treating sleep 
disorders in 
patients with 
schizophrenia. 
Zhao [5] They proposed an 
optimized restricted 
Boltzmann machine 
(RBM) feature 
extraction model 
combined with a 
softmax classification 
algorithm to detect 
electroencephalogram 
changes in children 
with autism spectrum 
disorder during music 
therapy. 
They 
validated the 
accuracy of 
the algorithm 
through 
experiments. 
Chen et al. 
[6] 
They investigated the 
effects of group music 
therapy on improving 
anxiety and self-
esteem in Chinese 
prisoners. 
The group 
music therapy 
appeared to be 
effective in 
improving 
anxiety and 
self-esteem, 
and was most 
beneficial for 
prisoners who 
were young or 
less educated.  
 

104 Informatica 47 (2023) 103 –108 D. Zhang  
 
 
 
3 Music therapy 
Routinely, anxiety gradually subsides without serious 
consequences as long as it is regulated by the autonomy of 
the psychological state, but there are also cases where 
stress or other factors lead to anxiety that cannot be 
relieved in time and eventually affects physiological 
health. In such cases, anxiety is no longer a conventional 
negative emotion [7], but rather a real psychological 
disease that requires external interventions for treatment. 
While medication serves as a natural form of anxiety 
treatment, it can only suppress the physiological 
symptoms presented and address the symptoms rather than 
the underlying cause; moreover, medication may also 
have side effects [8]. In addition to medication, music can 
also be used as a non-pharmacological approach to 
treating anxiety disorders [9]. Common music therapies 
include listening to music therapy, performing music 
therapy, and combined music and sports therapy. In 
listening to music therapy, the therapist selects appropriate 
music and musical rhythms to influence the patient's mood 
and behavior, resulting in stress reduction, pain relief, and 
improved sleep. On the other hand, performing music 
therapy allows patients to play instruments or sing to 
promote physical and mental health through musical 
activities. Combined music and exercise therapy allows 
patients to exercise with music to relax and adapt 
physically and mentally at the same time. 
4 Music emotion classification 
algorithm 
A brief introduction to music therapy was given earlier. 
The therapeutic principle of this treatment is to use the 
rhythmic melody and other characteristics of music to 
guide the relaxation of the patient's physical and mental 
state [10], thus interfering with negative psychological 
states. Regardless of the type of music therapy, the choice 
of music is very important in the process of treating 
patients. The therapist must select music that contains 
appropriate emotional states according to the patient's 
emotional state. For the therapist, an intelligent algorithm 
such as the music emotion classification algorithm can 
help therapists better understand the emotional 
characteristics of the music and thus make a more accurate 
selection. 
 
Input music 
video
Preprocessing
Feature 
extraction
Audio 
emotion 
classifier
Input music 
lyrics
Preprocessing
Text 
vectorization
Lyrics 
emotion 
classifier
Classification 
result fusion
 Music 
emotion 
classification 
results
 
Figure 1: Basic flow of multimodal fusion-based 
music emotion classification algorithm. 
 
Figure 1 shows the basic flow of the multimodal fusion-
based music emotion classification algorithm. In this 
paper, the Hevner emotion ring model is used for 
classifying music emotions, and there are eight emotion 
types. The specific steps of the algorithm are shown below. 
① The audio of the music and the corresponding lyrics 
are input respectively [11]. 
② The audio and lyrics are preprocessed separately. 
③ Features are extracted from the audio and lyrics using 
Mel-frequency cepstral coefficient (MFCC) and 
Word2vec, respectively [12]. The extraction formula of 
MFCC feature is: 
 

























= 




 +
=
=









+ 
+  
− − − − +
− +
  −
− − − − +
− −
− 
=








• =
=
 =




−
=
−
=
−
=
−
=
−
L l
M
m l
m S l c
k H
m f k
m f k m f
m f m f m f m f
k m f
m f k m f
m f m f m f m f
m f k
m f k
k H
k H P m S
k Y P
e n y k Y
M
m
M
m
m
m
N
k
m
N
n
N
kn j
, , 3 , 2 , 1
2
) 1 2 (
cos ) ( ) (
1 ) (
) 1 ( 0
) 1 ( ) (
)) 1 ( ) ( ))( 1 ( ) 1 ( (
) ) 1 ( ( 2
) ( ) 1 (
)) 1 ( ) ( ))( 1 ( ) 1 ( (
)) 1 ( ( 2
) 1 ( 0
) (
) ( ) ( ln ) (
) ( ) (
) ( ) (
1
1
1
0
1
0
2
1
0
2





,(1) 
 
where ) (k Y refers to the frequency domain signal after 
FFT [11], ) (n y refers to the original time domain signal, 
k refers to the number of sampling point, n
 
refers to the 
time sampling point of the time domain signal, 
) (  P
 is
 
the instantaneous energy of ) (k Y , ) (k H
m
 is the 
frequency response of a triangular filter, m is the serial 
number of a triangular filter in a group composing of M 
filters,  ) (m f
 
is the central frequency of the m -th 
triangular filter in a group, ) (l c is the L -order MFCC 
feature parameter, and ) (m S is the energy spectrum 
function of the frequency domain signal after filtering 
processing. 
④ An audio emotion classifier and a lyrics emotion 
classifier are used for emotion recognition, respectively. 
The audio emotion classifier performs emotion 
recognition on the Mel inverse spectrogram of the audio 
features, so a convolutional neural network (CNN) 
suitable for image recognition is used. When performing 
emotion recognition on audio, the CNN first performs 
convolutional computation on the Mel inverse 
spectrogram using convolutional kernels in the 
convolutional layer. It then compresses the convolutional 
feature map in the pooling layer. The formula of the 
convolutional computation is: 
 
) (
1 i i i i
B W O f O +  =
−
, (2) 
 
where 
i
O and 
1 − i
O are the feature maps output from the 
i -th and 1 − i -th layers, 
i
W is the weight in the i -th 
layer structure, 
i
B the bias in the i -th layer structure, and 
) (  f is the activation function. Convolution and pooling 

Study on the Effect of Decompression Music on Anxiety … Informatica 47 (2023) 103 –108 105 
 
 
 
in the CNN can be repeated as required, and the final 
convolutional feature maps will be computed in the fully 
connected layer to obtain the emotion recognition results. 
The lyrics emotion classifier performs emotion 
recognition on vectorized lyrics text using long short-term 
memory (LSTM) suitable for sequence information 
recognition [13]. 
⑤ After obtaining the emotion recognition results of 
the music using the audio emotion classifier and the lyrics 
emotion classifier [14], the results are weighted and fused 
according to the ratio of the weights that the two classifiers 
have. Finally, the overall emotion classification results of 
the music  are obtained. 
5 Example analysis 
5.1 Experimental environment 
Before testing the efficacy of music therapy using an 
emotion classification algorithm, the proposed 
multimodal fusion-based emotion classification algorithm 
was first tested to verify its effectiveness. The validation 
of the emotion classification algorithm was performed on 
a server in the laboratory. 
5.2 Experimental data 
Music and corresponding lyrics were crawled by a crawler 
program from the song lists of different emotion 
classifications on the NetEase Cloud platform for 
conducting the simulation experiments on the emotion 
classification algorithm. There were 12 emotion 
classifications in the song lists on the platform, and 50 
songs were crawled from each song list. In total, 600 
pieces of music and their corresponding lyrics were 
collected. The emotion classification of music in the 
NetEase Cloud platform is for user experience, and the 
emotion boundary between some song lists is relatively 
blurred, which is not conducive to the selection of music 
emotions in music therapy. Therefore, this paper used the 
Hevner emotion ring model to organize and summarize 
the crawled music. The music waveforms of some of these 
emotions are shown in Figure 2, from which it can be seen 
that there were obvious differences between the music 
waveforms of the two emotions. 
 
 
Figure 2: Music waveform diagrams of some emotions. 
5.3 Test items 
(1) Testing of a multimodal fusion-based emotion 
classification algorithm  
The parameters of the CNN as an audio emotion classifier 
in the multimodal fusion-based sentiment classification 
algorithm are shown below. The number of convolutional 
layers was set to 4, with 32 convolutional kernels of 2 2  
in each layer, and the sigmoid activation function was used. 
After every two convolutional layers, there was one 
pooling layer, and the mean pooling with a 3 3  pooling 
frame was used. The parameters of the LSTM as a lyrics 
emotion classifier are as follows. The number of 
dimensions of Word2vec vectorization was set to 200. 
Two hidden layers were used, with 512 nodes in each 
hidden layer, and the sigmoid function was used as the 
activation function. In addition to the above emotion 
classification algorithm, two unimodal emotion 
classification algorithms were tested as a comparison. 
(2) Testing of the efficacy of music therapy using emotion 
classification algorithms 
Eighty contemporaneous anxiety patients were selected 
from XX Hospital in Binzhou City, Shandong Province. 
All selected patients voluntarily participated in the test and 
signed an informed consent form. The patients were 
divided into two groups, the control group and the test 
group, and there was no significant difference between the 
two groups. The criteria for selection were meeting the 
diagnostic criteria for anxiety disorder, over 18 years old, 
being voluntarily participated in the test and signed an 
informed consent form. The exclusion criteria were 
intellectual disabilities, language barriers, severe systemic 
diseases, and severe organ dysfunction. 
Patients in both groups received conventional medication 
and conventional nursing interventions during the test 
process. The nursing interventions included education on 
medical knowledge related to anxiety disorders, rational 
use of medication, guidance on diet and rest, as well as the 
development and implementation of exercise training. The 
test group received additional music therapy, which was 
performed as follows. ① The therapist was required to 
conduct one-on-one treatment under the premise of 
gaining the patient's trust. During the treatment process, 
the therapist was required to fully understand the patient's 
condition and emotional characteristics. Before formally 
applying music therapy, the therapist was required to 
communicate with the patient for 15-20 minutes to guide 
the patient to express the negative emotions, providing 
part of the basis for the subsequent music selection. ② 
When music therapy was formally administered, the 
therapist selected the appropriate music for the patient 
based on the emotions previously expressed by the patient, 
combined with the emotion recognition results of the 
music to be selected by the emotion classification 
algorithm, and adjusted the patient's physical and mental 
state to a positive and optimistic state through music. 
The music therapy interventions for the patients were 
performed three times a week, with each session lasting 
one hour. The treatment lasted for six weeks in both the 
control and test groups. In addition, anxiety levels of 
patients in both groups were assessed before and after the 
treatment using the Self-Rating Anxiety Scale; psychiatric 
symptoms were assessed using the Psychiatric 
Assessment Scale. Finally, SPSS software [15] was used 

106 Informatica 47 (2023) 103 –108 D. Zhang  
 
 
 
for data statistics. The t-test was used to detect the 
differences between the two groups. 
5.4 Test results 
Figure 3 shows the emotion recognition accuracy test 
results of the three emotion classification algorithms. 
From Figure 3, it can be seen that the classification 
accuracy of the same emotion classification algorithm was 
similar for music with different emotions; for music with 
the same emotion, the multimodal fusion (CNN+LSTM)-
based emotion classification algorithm had the highest 
recognition accuracy among the three emotion 
classification algorithms, the LSTM-based algorithm for 
music lyrics was the second, and the CNN-based 
algorithm for music audio was slightly lower than the 
LSTM-based algorithm. 
 
 
Figure 3: Recognition accuracy of three emotion 
classification algorithms. 
 
Table 1 shows the anxiety scores of the control and test 
groups before and after the treatment. As can be seen from 
Table 1, before the treatment, the anxiety scores of both 
the control group and the test group were above 50, i.e., 
both suffered from anxiety disorders, and the difference 
between the two groups was not significant. After the 
treatment, the anxiety score of the control group decreased 
to below 50, and there was a significant difference 
compared to pre-treatment; the anxiety score of the test 
group decreased significantly compared to pre-treatment, 
their anxiety scores were below 50, i.e., the anxiety 
symptoms were significantly reduced, and the score was 
significantly lower compared to the control group. 
 
Table 1: Anxiety level of patients in both groups. 
 Before 
treatment 
After 
treatment 
P value 
Control 
group 
59.68 ± 
5.24 
49.68
 
± 
4.25 
0.002 
Test group 59.63 ± 
4.57 
37.21 ± 
2.56 
0.001 
P value 0.923 0.002  
 
Table 2 shows the results of the psychiatric assessment 
before and after the treatment in the control and test groups. 
As can be seen from Table 2, there was no significant 
difference in the scores on the same psychiatric 
assessment dimensions between the control and test 
groups before the treatment, and both maintained 
relatively high scores. After the treatment, there was a 
significant decrease in the scores of the control and test 
groups on all five assessment dimensions, indicating that 
the treatment modalities received by both groups were 
effective. When comparing the scores of the two groups 
on the same dimensions, it can be seen that the score of 
the test group was significantly lower. 
 
Table 2: Results of psychiatric assessment in both 
groups before and after treatment. 
Psychiat
ric 
assessm
ent 
dimensi
ons 
Group Before 
treatmen
t 
After 
treatmen
t 
P value 
Anxiety 
and 
depressi
on 
Control 
group 
6.13 ± 
1.11 
5.03 ± 
1.13 
0.000 
Test 
group 
6.11 ± 
1.02 
3.25 ± 
1.14 
0.001 
P value 0.689 0.000  
Thought 
disorder
s 
Control 
group 
6.89 ± 
1.24 
5.12 ± 
1.11 
0.002 
Test 
group 
6.78 ± 
1.13 
3.14 ± 
1.47 
0.000 
P value 0.788 0.001  
Lack of 
vitality 
Control 
group 
6.32 ± 
1.41 
5.11 ± 
1.23 
0.001 
Test 
group 
6.22 ± 
1.36 
3.05 ± 
1.57 
0.000 
P value 0.874 0.001  
Activati
on 
Control 
group 
6.35 ± 
1.53 
5.24 ± 
1.63 
0.002 
Test 
group 
6.33 ± 
1.44 
3.07 ± 
1.21 
0.001 
P value 0.845 0.000  
Hostility 
and 
suspicio
n 
Control 
group 
6.58 ± 
1.11 
5.16 ± 
1.25 
0.001 
Test 
group 
6.55 ± 
1.37 
3.11 ± 
1.14 
0.000 
P value 0.748 0.001  
6 Discussion 
Under the pressure of fast-paced life, people tend to be 
more prone to anxiety. In general, anxiety can be relieved 
through psychological self-regulation, but sometimes 
anxiety cannot be relieved in time due to stress or other 
factors and eventually induces a mental illness such as 
anxiety disorder, which interferes with normal life. In this 
case, it is necessary to intervene using external means. 
Medication is a common treatment, but it is not a cure and 
has side effects. Music can also be used to treat anxiety 
disorders. In the process of using music therapy, the 
therapist must choose music that contains the right 
emotions according to the patient's current mood in order 
to regulate their negative emotions. Therefore, it is 

Study on the Effect of Decompression Music on Anxiety … Informatica 47 (2023) 103 –108 107 
 
 
 
important to accurately identify and select the emotions 
contained in the music. In this paper, a multimodal fusion-
based algorithm for emotion classification was used to 
identify the emotions in music and assist therapists in 
selecting the music with appropriate emotion. In the 
subsequent experiments, the performance of the 
multimodal fusion-based emotion classification algorithm 
was verified by comparing it with the CNN-based audio 
emotion classification algorithm and the LSTM-based 
lyric emotion algorithm. Multimodal fusion recognized 
music emotion as a whole from both audio and lyrics 
perspectives, so it had a better recognition performance 
than the other two one-sided algorithms. 
After validating the accuracy of the emotion classification 
algorithm, the efficacy of music therapy was tested on 80 
patients with anxiety. They were divided into a control 
group and a test group, both of which received the same 
conventional medication and care during the treatment 
cycle. However, only the test group received additional 
music therapy. Anxiety and psychiatric assessments were 
conducted before and after treatment in both groups. The 
final results showed that conventional medication and care 
were effective in relieving anxiety, but the test group 
receiving music therapy had better anxiety relief and more 
significant results. The reasons are as follows. 
Conventional medication and nursing care can suppress 
the surface anxiety symptoms, and the requirement for 
sufficient exercise during the care process facilitated 
adjustment of psychological state through physiological 
changes. Therefore, the control group, which only 
received conventional treatment, experienced significant 
relief of anxiety after treatment. In the test group, music 
therapy was added. The audio and rhythm of music could 
resonate with the body cells of patients and stimulate their 
cerebral cortex to indirectly regulate and control emotions. 
At the same time, music could also divert patients' 
attention away from negative thoughts and relieving their 
anxiety at a psychological level. 
Compared to the studies mentioned earlier regarding 
music therapy, which mostly focused on examining the 
therapeutic effects of music therapy on mental disorders 
such as depression and anxiety, this article instead 
emphasized how to select appropriate music for treating 
anxiety. When classifying and identifying the emotions 
contained in music, it is not like traditional recognition 
methods that only distinguish the emotions of music based 
on audio or lyrical text. Instead, it combined the audio and 
lyrical text of the music, using CNN to recognize the audio 
and LSTM to recognize the lyrical text. Then, a weighted 
combination of both recognition results was used to obtain 
the final result. This paper utilized multimodal features 
such as audio and text to classify and recognize the 
emotions in music, thereby providing appropriate 
emotional music tailored to patients' actual experiences. 
Furthermore, this study examined the effectiveness of the 
proposed algorithm in music therapy specifically for 
anxiety patients, offering valuable insights for enhancing 
the efficacy of music therapy. 
The limitation of this study lies in the insufficient diversity 
and quantity of samples used to train the emotion 
classification algorithm, which can easily lead to 
overfitting. Therefore, future research directions include 
expanding the range and number of samples to further 
enhance the recognition performance of the emotion 
classification algorithm for music, thus better assisting 
music therapy. 
7 Conclusion 
This paper briefly introduces music therapy and the 
multimodal fusion-based emotion classification algorithm 
used to accurately select music with appropriate emotions 
for music therapy. Then, the multimodal fusion-based 
emotion classification algorithm was tested in an 
experiment, and the efficacy of music therapy was tested 
on 80 contemporaneous anxiety patients from XX 
Hospital in Binzhou City, Shandong Province.The 
following results were obtained. (1) The multimodal 
fusion-based (CNN+LSTM-based) emotion classification 
algorithm had the highest recognition accuracy, the 
LSTM-based music lyrics emotion classification 
algorithm was the second, and the CNN-based music 
audio emotion classification algorithm was slightly lower 
than the LSTM-based algorithm. (2) Both the control and 
test groups showed significant relief from anxiety after 
treatment, and the test group that additionally received 
music therapy showed greater relief. (3) After the 
treatment, psychiatric scores were significantly reduced in 
both the control and test groups, with a more significant 
reduction in the test group. 
8 References 
[1] Eseadi C, Ngwu M O (2023). Significance of music 
therapy in treating depression and anxiety disorders 
among people with cancer. World Journal of 
Clinical Oncology: English Edition, 14, pp. 69-80. 
https://doi.org/10.5306/wjco.v14.i2.69 
[2] Mayer-Ferbas J, Gassner L (2021). PP90 
Effectiveness of Music Therapy for Autism 
Spectrum Disorder, Dementia, Depression, Insomnia, 
And Schizophrenia. International Journal of 
Technology Assessment in Health Care, 37, pp. 18-
18. https://doi.org/10.1017/S0266462321001094 
[3] Chow R, Bergner R, Prsic E (2021). Music therapy 
effectiveness by duration in patients with cancer: a 
meta-regression. Supportive and Palliative Care, 13, 
pp. 117-120. https://doi.org/10.1136/bmjspcare-
2021-003163 
[4] Lu M J, Chen W Y, Li D J (2022). Efficacy of music 
therapy and predictors of sleep disturbance among 
patients with chronic schizophrenia: A prospective 
study. Archives of Psychiatric Nursing, 40, pp. 1-7. 
https://doi.org/10.1016/j.apnu.2022.04.001 
[5] Zhao N (2021). Intelligent System of Somatosensory 
Music Therapy Information Feedback in Deep 
Learning Environment. Complexity, 2021, pp. 1-11. 
https://doi.org/10.1155/2021/9935504 
[6] Chen X J, Hannibal N, Gold C (2016). Randomized 
Trial of Group Music Therapy with Chinese 

108 Informatica 47 (2023) 103 –108 D. Zhang  
 
 
 
Prisoners: Impact on Anxiety, Depression, and Self-
Esteem. International Journal of Offender Therapy 
& Comparative Criminology, 60, pp. 1064-1081. 
https://doi.org/10.1177/0306624X15572795 
[7] Monsalve-Duarte S, Betancourt-Zapata W, Suarez-
Canon N, Maya R, Salgado-Vasco A, Prieto-Garces 
S, Marín-Sánchez J, Gómez-Ortega V, Valderrama, 
Ettenberger M (2022). Music therapy and music 
medicine interventions with adult burn patients: A 
systematic review and meta-analysis. Burns: 
Including Thermal Injury, 48, pp. 510-521. 
https://doi.org/10.1016/j.burns.2021.11.002 
[8] Sun B, Tian F, Jia M (2019). Emotion recognition 
method of Tibetan micro-blog text based on 
sentiment dictionary. Journal of Physics: 
Conference Series, 1314, pp. 1-6. 
https://doi.org/10.1088/1742-6596/1314/1/012182 
[9] Adarsh S R (2019). Enhancement of text-based 
emotion recognition performances using word 
clusters. International Journal of Research - 
GRANTHAALAYAH, 7, pp. 238-250. 
https://doi.org/10.29121/granthaalayah.v7.i1.2019.1
051 
[10] Thakur P, Shrivastava R (2018). A Review on Text 
Based Emotion Recognition System. International 
Journal of Advanced Trends in Computer Science 
and Engineering, 7, pp. 67-71. 
https://doi.org/10.30534/ijatcse/2018/01752018 
[11] Gao Y, Xu A, Hu J H, Cheng T H (2017). 
Incorporating association rule networks in feature 
category-weighted naive Bayes model to support 
weaning decision making. Decision Support Systems, 
96, pp. 27-38. 
https://doi.org/10.1016/j.dss.2017.01.007 
[12] Xiao S, Hu Y, Han J, Zhou R, Wen J (2016). 
Bayesian Networks-based Association Rules and 
Knowledge Reuse in Maintenance Decision-Making 
of Industrial Product-Service Systems. Procedia 
CIRP, 47, pp. 198-203. 
https://doi.org/10.1016/j.procir.2016.03.046 
[13] Rao W, Zhu L, Pan S, Yang P, Qiao J (2019). 
Bayesian Network and Association Rules-based 
Transformer Oil Temperature Prediction. Journal of 
Physics Conference, 1314, pp. 1-8. 
https://doi.org/10.1088/1742-6596/1314/1/012066 
[14] Alswaidan N, Menai M (2020). Hybrid Feature 
Model for Emotion Recognition in Arabic Text. 
IEEE Access, 4, pp. 1-12. 
https://doi.org/10.1109/ACCESS.2020.2975906 
[15] Kiruthika S, Priyan V V (2021). Detection of Online 
Fake Reviews Based on Text Emotion. Journal of 
Physics: Conference Series, 1916, pp. 1-8. 
https://doi.org/10.1088/1742-6596/1916/1/012153