=Paper=
{{Paper
|id=Vol-3137/paper9
|storemode=property
|title=Method of Forming the Context of Advertising and Target Audience based on Associative Rules Learning
|pdfUrl=https://ceur-ws.org/Vol-3137/paper9.pdf
|volume=Vol-3137
|authors=Lipianina-Honcharenko Khrystyna,Taras Lendiuk,Anatoliy Sachenko,Jacek Woloszyn
|dblpUrl=https://dblp.org/rec/conf/cmis/KhrystynaLSW22
}}
==Method of Forming the Context of Advertising and Target Audience based on Associative Rules Learning==
https://ceur-ws.org/Vol-3137/paper9.pdf
Method of Forming the Context of Advertising and Target
Audience based on Associative Rules Learning
Khrystyna Lipianina-Honcharenko 1, Taras Lendiuk 1, Anatoliy Sachenko 1,2, Jacek
Wołoszyn 2
1
West Ukrainian National University, Lvivska Str., 11, Ternopil, 46000, Ukraine
2
Kazimierz Pulaski University of Technology and Humanities in Radom, ul.Malczewskiego 29, 26-600 Radom,
Poland
Abstract
Nowadays, important mechanisms of study are content and techniques of its creation, the
problem of influencing the target audience, which itself seeks to shape communication
processes. Internet content occupies a position of powerful communication technology, which
continues to grow rapidly and gain influence. Creating a large number of advertisements,
especially texts, is extremely expensive. Therefore, it is worth considering how generate these
texts automatically. In this regard, it is possible to assume that the development of a method of
forming the context of advertising and target audience based on learning associative rules is
relevant and can increase the effectiveness of advertising, and thus reduce the cost of online
advertising of higher education institutions. The input data used a survey of students majoring
in Computer Science, regarding admission. The 152 students took part in the survey and
answered 10 questions. The experimental results confirmed, the proposed method enabled to
increase the effectiveness of advertising on social networks at least in 23%, and reduce the
price in 90%.
Keywords 1
Data analysis, advertising content, Associative Rules Learning, Apriori algorithm, Facebook.
1. Introduction
The value of advertising is crucial for a company, because only this can make people aware of the
company's product and, doing so, can create a good opportunity to sell it to customers [2]. The manual
work of a large number of advertisements, especially texts, is extremely expensive. Therefore, it is
worth considering how to generate these texts automatically.
Therefore, we can assume that the development of a method of forming the context of advertising
and target audience based on learning associative rules is relevant and can increase the effectiveness of
advertising, and thus reduce the cost of online advertising of higher education institutions.
This work is devoted to this topic and it is distributed as follows. Section 2 discusses the analysis of
related work; section 3 presents a method of forming the context of advertising and the target audience
based on the associative rules learning. Section 4 presents the implementation of the method. Section 5
presents the conclusions of the study.
2. Related Work
Given the complexity of the modern digital advertising ecosystem, there are many studies describing
the impact of advertising content on social networks on attracting customers [19], using data from
Facebook in: medicine [20], psychology [21], sociology [22], politics [23] and others. Research [24,
CMIS-2022: The Fifth International Workshop on Computer Modeling and Intelligent Systems, Zaporizhzhia, Ukraine, May 12, 2022
EMAIL: xrustya.com@gmail.com (Kh. Lipyanina-Goncharenko); tl@wunu.edu.ua (T. Lendiuk); as@wunu.edu.ua (A. Sachenko);
jacek.woloszyn@uthrad.pl (J. Wołoszyn)
ORCID: 0000-0002-2441-6292 (Kh. Lipyanina-Goncharenko); 0000-0001-9484-8333 (T. Lendiuk); 0000-0002-0907-3682 (A. Sachenko)
© 2022 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
25] provides an understanding of social networks for higher education institutions such as Instagram,
Pinterest, Snapchat and WhatsApp.
In [13], a model is proposed that is able to use its targeting strategy in accordance with the received
feedback. The model uses the Thompson Sampling algorithm applied to the user space. [1] a set of
models of regression, cluster analysis and analysis of association rules to find patterns of user behavior
in relation to marketing campaigns, taking into account user characteristics and financially significant
indicators.
The aim of the article [12] is to study the analysis of social media data using machine learning tools;
new approach to social media marketing strategy uses the Waikato Knowledge Analysis Environment
(WEKA). In [5] the article implemented the level of the aspect of mood analysis, based on machine
learning algorithms of SVM and NB classification.
In [17] the study analyzed the various needs of customers, focusing on online advertising based on
methods of classification, segmentation and clustering. In [16] the model of selection of the optimal
amount of advertising on various Internet resources is analyzed in order to achieve the desired coverage
of the target audience. In addition, the method of multi-criteria optimization with the definition of the
obtained objective function is considered, which allows considering simultaneously the various aspects
of media selection issue and optimal budgeting and budget allocation. The proposed approach [2] draws
on knowledge that can support several solutions, ranging from marketing campaigns for each customer
segment, redesign of the store layout to product recommendations. In [15] the development of
advertising art design on the basis of information technologies is mainly investigated.
In a study [4, 26], it is proposed to apply the training of association rules to find influential bloggers
in time using the Apriori algorithm. An improved Apriori algorithm has been proposed to identify the
relationship between the TECP and the thirty-five factors, which cover four categories of household
characteristics, including housing, socio-demographic, household appliances and heating, and energy
attitudes [9]. In [10] a new effective system of recommendations based on the Apriori algorithm for
user requirements was proposed.
Article [18] proposes the generation of advertising texts based on keywords that take into account
product information. Ins [3] was developed a web-based system of recommendations for choosing a
property using the method of content-based filtering. The referral system provides information about
properties based on user behavior by searching for advertising content that the user previously searched
for. [11] presents an intelligent management system for advertising on social networks, based on data
analysis techniques to automatically create ads.
The above-mentioned works mostly analyze the actions of users on online advertising. There are
also a number of works that present research on methods of associative rules learning. There are also
works that use different approaches to the formation of content and target audience of Internet
advertising (analogues).
Therefore, goal of this paper is to develop a method of forming the context of advertising and the
target audience based on the associative rules learning.
The novelty of the work is the formation of the most profitable (in the economic aspect) text of the
Free Economic Zone advertising and the corresponding target group, which will increase the
effectiveness of the advertising campaign based on learning the rules of association.
3. Proposed Method
To reduce the time spent on the formation of advertising content for the target audience, the authors
have developed a method of forming the context of advertising and the target audience based on the
associative rules learning. The proposed method is illustrated schematically (Fig. 1) and is represented
by the following steps:
Step 1. Conduct a survey of students (Block 1). It is important to determine the gender characteristics
of the respondents, as this will make it possible to determine the target audience in the future. Convert
to csv format (Block 2).
1 2
Survey
Formatting in csv.
of students
4 3
Convert data Definition
to a list of support
Learning associative rules 5
method Apriori
Search 5.1
for 5.2
Remove pairs of rare
pairs of
items
elements
5.3
Forming rules
Element X => Element Y
5.4 5.5
Calculating
Lift calculation
confidence
6 7 Formation
Derivation of
context of advertising
rules
and target audience
Figure 1: Algorithmic structure of forming the context of advertising and target audience based on
learning associative rules
Step 2. Calculation of support for each individual element (Block 3). Support is simply the number
of transactions during which a particular product (or combination of products) occurs.
Step 3. Conversion of data into a list (Block 4).
Step 4. Start (Block 5) learning associative rules based on the Apriori method [6].
Step 4.1. Finding support for frequent sets of elements (Block 5.1). Search for pairs of items that
appear most often in pairs. The Apriori algorithm ignores all pairs that contain any of the rare elements
(Block 5.2).
Step 4.2. Formation of rules (Block 5.3). The most frequent sets of elements, converted into
association rules, in the format: Element X => Element Y.
Step 4.3. Calculation of confidence (Block 5.4). Confidence shows the percentage of cases in which
this rule applies.
Step 4.3. Elevator calculation (Block 5.5). Raising a rule is an indicator of effectiveness, which
indicates the strength of the relationship between the products in the rule. Raising the rule is determined
by the following formula:
𝑃 𝑋∩𝑌
𝑙𝑖𝑓𝑡
𝑃 𝑋 𝑃 𝑌
where, P is the probability of the frequency of combination of elements in the generated rule.
Step 5. Output of results-rules (Block 6).
Step 6. Creating the context of advertising and target audience (Block 7), for higher education
institutions on the basis of the received rules.
4. Experimental Results
The Python language was chosen to conduct the method of forming the context of advertising and
the target audience based on th associative rules learning.
The input data used a survey of students majoring in Computer Science, regarding admission. 152
students took part in the survey and answered 10 questions. All students’ feedbacks are in .csv format.
The sample is representative, as the survey was conducted among students majoring in Computer
Science, who themselves have recently been faced with the choice of where to enter, so their feedback
is the most informative for this type of advertising.
Before starting the analysis of the rules of the association, first determine the frequency distribution
of the elements (Table 2). The chart shows that the majority of male respondents, as well as the highest
number of answers, said that they found information on their own in social networks, majoring in
Computer Science.
Table 1
Response rate
Item name Count
Male 67
independently found information 59
I saw a lot of interesting information on social networks 41
advised familiar relatives 39
Female 37
Interest in computers 35
The presentation of the specialty by the representatives was interesting 33
Interest in design 32
Interest in technology 30
Passed here for public training 29
Augmented reality 29
I got a call and was convinced that it would be interesting to study here 29
Representatives of the specialty came to our school 20
Internet of Things 17
Robotics 15
Internet of Things. Augmented reality 12
Software control of drones 11
The cost of training suited 11
From social networks 10
Due to quarantine, I decided to choose a place closer to home 8
Interest in IT 6
Internet of Things. Augmented reality. Robotics. Software control of drones 5
Augmented reality. Robotics. 4
Internet of Things. Robotics. 4
Augmented reality. Software control of drones 4
Augmented reality. Robotics. Software control of drones 2
Internet of Things. Augmented reality. Robotics. 2
Item name Count
Internet of Things. Robotics. Software control of drones 2
Quality presentation of the material 2
A friend studied in the same specialty 1
A friend recommended 1
Web 1
I wanted to study to be a programmer 1
Because a large variety of specialties could not be determined 1
Robotics. Software control of drones 1
Chose occasionally 1
I like working with computers and I am interested in the specialty 1
Prospect 1
My chosen specialty is the most relevant 1
Because this specialty covers many areas that I really like 1
Many acquaintances study in Ternopil 1
A relative studied at the university 1
I have wanted to become a programmer for a long time, so I chose this specialty 1
I found information from social networks on my own 1
I really wanted to enroll in FCIT and the most advanced training is on CS 1
I was interested in computer science 1
In addition, it is important to note that even the most frequent response of more than 11% is male
(Table 2). Next, we will use this information as a guide when setting a minimum support threshold.
Table 2
The most common answers
Index Item Count Percentage
13 augmented reality 29 4.93
40 passed here for public training 29 4.93
47 interest in technology 30 5.10
45 interest in design 32 5.44
37 interested in the presentation of the specialty by 34 5.78
representatives
46 interest in computers 35 5.95
34 female 37 6.29
39 advised familiar relatives 39 6.63
31 saw a lot of interesting information on social networks 41 6.97
41 independently found information 59 10.03
43 male 67 11.39
After converting the dataset into the desired list, we will display the results, namely the rules, which
in the future will allow to form the context of advertising and target groups.
Therefore, after starting the algorithm was generated:
1. Counting sets of items of length 1:
48 candidates were found for sets of length 1;
Found 15 large sets of items of length 1.
2. Counting sets of items of length 2:
105 candidates were found for sets of length 2;
32 large sets of items of length 2 were found.
Based on the experimentally significant parameters of the algorithm, the generated rules are filtered.
Parameters:
min_support = 0.11 – defined as the percentage of most frequent responses;
min_confidence = 0.65 – this probability of determining the rules is sufficient and in experimental
studies showed quite good results for this sample.
Rules:
{male, interest in design} -> {augmented reality} (conf: 0.737, supp: 0.135, lift: 1.965, conv:
2.375);
{advised by familiar relatives, passed here for public education} -> {male} (conf: 1.000, supp:
0.115, lift: 1.552, conv: 355769230.769);
{saw a lot of interesting information on social networks, advised familiar relatives} -> {male}
(conf: 0.857, supp: 0.115, lift: 1.330, conv: 2.490);
{saw a lot of interesting information on social networks, interest in technology} -> {found
information on his own} (conf: 0.706, supp: 0.115, lift: 1.244, conv: 1.471);
{independently found information, interest in computers} -> {male} (conf: 0.800, supp: 0.154,
lift: 1.242, conv: 1.779);
{interest in technology} -> {independently found information} (conf: 0.700, supp: 0.202, lift:
1.234, conv: 1.442);
{passed here for public training} -> {male} (conf: 0.793, supp: 0.221, lift: 1.231, conv: 1.720);
{advised familiar relatives, interest in design} -> {male} (conf: 0.778, supp: 0.135, lift: 1.207,
conv: 1.601);
{interest in computers} -> {male} (conf: 0.771, supp: 0.260, lift: 1.197, conv: 1.556);
{Augmented Reality} -> {male} (conf: 0.690, supp: 0.192, lift: 1.071, conv: 1.146);
{saw a lot of interesting information on social networks} -> {male} (conf: 0.659, supp: 0.260,
lift: 1.022, conv: 1.042).
Almost all of the received rules mention the gender of the respondent (male), which indicates the
main target audience for the specialty “Computer Science”. Rules that are more than two elements of
the answer, allow to create content for advertising. Let's form some examples of content (Table 3).
Table 3
Formation of advertising content in relation to the generated rules
# of variant Rule Content
1 {male, interest in design} ‐> {Augmented Interested in Design, Try Yourself in
reality} Augmented Reality
2 {advised by familiar relatives, passed We Are Recommended When Public
here for public education} ‐> {male} Education Is Important
3 {saw a lot of interesting information on We are recommended after browsing
social networks, advised familiar our social networks
relatives} ‐> {male}
4 {saw a lot of interesting information on Interested in technology, visit our social
social networks, interest in technology} ‐ networking pages, there's lots of
> {male} interesting information
5 {self‐found information, interest in Interested in Technology, Visit our Social
computers} ‐> {male} Networking Pages
To compare the effectiveness of the generated advertising content based on the associative rules
learning, a comparative experiment was conducted on Facebook on the business page “Computer
Science of ZUNU”. The first option (Option 0) advertising (Fig. 2), developed on the basis of the rules
identified in previous research, namely:
The greatest interaction with the video advertising of Facebook “Computer Science ZUNU”
had males in the age category 18-25, 35-55 [7];
Male and female clients in the age category of 40-55 had the greatest interaction with the ZUNU
Computer Science business page.
Variant 0
Figure 2: Previous version of the advertisement “Computer Science” of the Western Ukrainian
National University on Facebook
Fig. 3 presents an advertisement formed on the basis of learning associative rules, a comparative
experiment was conducted on Facebook on the business page “Computer Science of ZUNU”. The
content used is from Table 3 and the target audience is male for all age groups.
Variant 1 Variant 2 Variant 3
Variant 4 Variant 5
Figure 3: New variants of advertising "Computer Science" of the Western Ukrainian National
University on Facebook
Table 4 presents a comparison of the effectiveness of the generated advertising content based on the
associative rules learning, in the period from May 4, 2021 – May 31, 2021 with all content options,
including the old.
Table 4
Comparison of the effectiveness of the generated advertising content
Advertising Results Coverage Readings Price for the
variant result
Index Changes Index Changes Index Changes Index Changes
Variant 0 120 100% 4498 100% 5395 100% 0,23 100%
Variant 1 147 123% 6561 146% 8776 163% 0,08 37%
Variant 2 197 164% 6364 141% 6560 122% 0,06 25%
Variant 3 240 200% 8561 190% 8192 152% 0,02 10%
Variant 4 160 133% 5442 121% 6671 124% 0,12 53%
Variant 5 196 163% 7024 156% 8867 164% 0,03 14%
Table 4 shows that all variants of the new ad gave improved results. The result indicator shows how
many times customers have been in contact with the ad. Advertising performed best in Option 3 (Fig. 4),
as it performed 100% better than Option 0. Option 3 performed best in all indicators. Namely, the
coverage is 90% better than option 0 and 52% better in terms of the number of impressions. It also
reduced the price for the result by 90% in option 3.
2,5 1,2
2 1
0,8
1,5
0,6
1
0,4
0,5 0,2
0 0
Variant 0 Variant 1 Variant 2 Variant 3 Variant 4 Variant 5
Results Coverage Readings Price for the result
Figure 4: Comparison of the effectiveness of the generated advertising content
Thus, the proposed method enabled to increase the effectiveness of advertising on social networks
at least in 23%, and reduce the price in 90%. Moreover, the authors believe that growing the number of
student surveys can increase the quality of associative rules learning as well as get better keywords for
content formation. That will lead to increasing the effectiveness of advertising and reducing its costs
respectively.
5. Conclusions
Developed method of forming the context of advertising and target audience based on the associative
rules learning makes it possible to increase the effectiveness of advertising, and thus reduce the cost of
online advertising of higher education institutions. Also, the developed method will allow to form rules
between the answers of respondents for the formation of advertising content and the definition of the
target group.
To implement the developed method, we used a survey of students majoring in Computer Science,
regarding admission. 152 students took part in the survey and answered 10 questions.
The proposed method enabled to increase the effectiveness of advertising on social networks at least
in 23%, and reduce the price in 90%. Moreover, the authors believe that growing the number of student
surveys can increase the quality of associative rules learning as well as get better keywords for content
formation. That will lead to increasing the effectiveness of advertising and reducing its costs
respectively.
Unlike analogues [3, 11, 18] the developed a method enables to form rules between the answers of
respondents, construct the advertising content and determine the target group.
In the future, the authors plan to develop the information system that will automatically generate the
content of the advertising message and select the target audience using the deep learning methods
[27, 28].
6. References
[1] M. M. Monastyrskaya, V. I. Soloviev, Improving customer relationship management based on
intelligent analysis of user behavior patterns, in: Proceedings of the 2020 13th International
Conference “Management of Large-Scale System Development” (MLSD), 2020, pp. 1-
4. doi:10.1109/mlsd49919.2020.92477 doi: 10.1109/MLSD49919.2020.9247718.
[2] A. Griva, C. Bardaki, K. Pramatari, Katerina, D. Papakyriakopoulos, Retail business analytics:
Customer visit segmentation using market basket data, Expert Systems with Applications, 100
(2018), 1-16. doi: 10.1016/j.eswa.2018.01.029.
[3] T. Badriyah, S. Azvy, W. Yuwono, I. Syarif, Recommendation system for property search using
content based filtering method, in: Proceedings of the 2018 IEEE International Conference on
Information and Communications Technology (ICOIACT), 2018, 25–29. doi:
10.1109/ICOIACT.2018.8350801.
[4] B. Shazad, H. U. Khan, M. Farooq, A. Mahmood, I. Mehmood, S. Rho, Y. Nam, Finding temporal
influential users in social media using association rule learning, Intelligent Automation And Soft
Computing 26 (2020) 87–98. doi: 10.31209/2019.100000130.
[5] S. Vanaja, M. Belwal, Aspect-level sentiment analysis on e-commerce data, in: Proceedings of the
2018 International Conference on Inventive Research in Computing Applications (ICIRCA), 2018,
pp. 1275-1279. doi: 10.1109/ICIRCA.2018.8597286.
[6] R. Agrawal, R. Srikant, Fast algorithms for mining association rules, in: Proceedings of the 20th
International Conference on Very Large Data Bases VLDB, September 1994, vol. 1215, pp. 487-
499.
[7] H. Lipyanina, S. Sachenko, T. Lendyuk, A. Sachenko, Targeting model of HEI video marketing
based on classification tree, in: Proceedings of the 16th International Conference on ICT in
Education, Research and Industrial Applications. Integration, Harmonization and Knowledge
Transfer. Volume II: Workshops, ICTERI 2020, Kharkiv, Ukraine, 6-10 October 2020, CEUR
Workshop Proceedings, vol. 2732, pp. 487-498. http://ceur-ws.org/Vol-2732/20200487.pdf.
[8] H. Lipyanina, A. Sachenko, T. Lendyuk, S. Nadvynychny, S. Grodskyi, Decision tree based
targeting model of customer interaction with business page, in: Proceedings of the third
International Workshop on Computer Modeling and Intelligent Systems (CMIS-2020), April 27 –
May 1, 2020. CEUR Workshop Proceedings, vol. 2608, pp. 1001–1012. http://ceur-ws.org/Vol-
2608/paper75.pdf.
[9] F. Wang. K. Li. N. Duić. Z. Mi. B.-M. Hodge. M. Shafie-khah. J. P. S. Catalão, Association rule
mining based quantitative analysis approach of household characteristics impacts on residential
electricity consumption patterns, Energy Conversion and Management 171 (2018) 839–854. doi:
10.1016/j.enconman.2018.06.017.
[10] S. Alzu'bi, B. Hawashin, M. Eibes and M. Al-Ayyoub, A novel recommender system based on
apriori algorithm for requirements engineering, in: Proceedings of the 2018 Fifth International
Conference on Social Networks Analysis, Management and Security (SNAMS), 2018, pp. 323-
327. doi: 10.1109/SNAMS.2018.8554909.
[11] J. Aguilar, G. Garcia, An adaptive intelligent management system of advertising for social
networks: A case study of facebook, IEEE Transactions on Computational Social Systems 5 (2018)
20-32. doi: 10.1109/TCSS.2017.2759188.
[12] B. S. Arasu, B. J. B. Seelan, N. Thamaraiselvan, A machine learning-based approach to enhancing
social media marketing, Computers & Electrical Engineering 86 (2020) 106723. doi:
10.1016/j.compeleceng.2020.106723.
[13] A. Popov, D. Iakovleva, Adaptive look-alike targeting in social networks advertising, Procedia
computer science 136 (2018) 255-264. doi: 10.1016/j.procs.2018.08.264.
[14] N. Shah, S. Engineer, N. Bhagat, et al., Research trends on the usage of machine learning and
artificial intelligence in advertising, Augment Hum Res 5 (2020). doi: 10.1007/s41133-020-00038-
8.
[15] Z. Liu, Development of advertising art design based on information technology, in: J. Jansen B.,
Liang H., Ye J. (eds) International Conference on Cognitive based Information Processing and
Applications (CIPA 2021), volume 85 of Lecture Notes on Data Engineering and Communications
Technologies, Springer, Singapore, 2021, pp. 3-10. doi: 10.1007/978-981-16-5854-9_1.
[16] O. Barabash, G. Shevchenko, N. Dakhno, O. Neshcheret, A. Musienko, Information technology of
targeting: optimization of decision making process in a competitive environment, International
Journal of Intelligent Systems and Applications 9 (2017) 1-9. doi: 10.5815/ijisa.2017.12.01.
[17] R. Saito, K. Otake, T. Namatame, Analysis of fashion market trend using advertising data of
shopping information site, in: Meiselwitz G. (eds) Social Computing and Social Media.
Participation, User Experience, Consumer Experience, and Applications of Social Computing.
HCII 2020, volume 12195 of Lecture Notes in Computer Science, Springer, Cham, 2020, pp. 389-
400. doi: 10.1007/978-3-030-49576-3_28.
[18] K. Wakimoto, S. Kawamoto, P. Zhang, Keyword-based text generation for internet advertisement,
in: Proceedings of the 34th Annual Conference of the Japanese Society for Artificial Intelligence,
2020, pp. 1-4.
[19] D. Lee, K. Hosanagar, H. S. Nair, Advertising content and consumer engagement on social media:
Evidence from Facebook, Management Science 64 (2018) 5105-5131. doi:
10.1287/mnsc.2017.2902.
[20] A. M. Jamison, D. A. Broniatowski, M. Dredze, Z. Wood-Doughty, D. A. Khan, S. C. Quinn,
Vaccine-related advertising in the Facebook Ad Archive, Vaccine 38, (2019) 512-520. doi:
10.1016/j.vaccine.2019.10.066.
[21] S. Youn, S. Kim, Understanding ad avoidance on Facebook: Antecedents and outcomes of
psychological reactance, Computers in Human Behavior 98 (2019) 232-244. doi:
10.1016/j.chb.2019.04.025.
[22] C. L. White, B. Boatwright, Social media ethics in the data economy: Issues of social responsibility
for using Facebook for public relations, Public Relations Review 46, (2020) 101980. doi:
10.1016/j.pubrev.2020.101980.
[23] A. Gitomer, P. V. Oleinikov, L. M. Baum, et al., Geographic impressions in Facebook political
ads, Appl Netw Sci 6, (2021). doi: 10.1007/s41109-020-00350-7.
[24] A. Peruta, A, B. Shields, Social media in higher education: understanding how colleges and
universities use Facebook, Journal of Marketing for Higher Education 27 (2017) 131-143. doi:
10.1080/08841241.2016.1212451.
[25] S. Manca, Snapping, pinning, liking or texting: Investigating social media in higher education
beyond Facebook, The Internet and Higher Education 44 (2020) 100707. doi:
10.1016/j.iheduc.2019.100707.
[26] A. Alhegami, H. Alsaeedi, A framework for incremental parallel mining of interesting association
patterns for big data, International Journal of Computing 19 (2020) 106-117. URL:
http://computingonline.net/computing/article/view/1699
[27] M. Komar, A. Sachenko, V. Golovko and V. Dorosh, Compression of network traffic parameters
for detecting cyber attacks based on deep learning, in: Proceedings of the 2018 IEEE 9th
International Conference on Dependable Systems, Services and Technologies (DESSERT), 2018,
pp. 43-47, doi: 10.1109/DESSERT.2018.8409096.
[28] Zh. Hu, Ye. V. Bodyanskiy, N. Ye. Kulishova, O. K. Tyshchenko, A multidimensional extended
neo-fuzzy neuron for facial expression recognition, International Journal of Intelligent Systems
and Applications (IJISA) 9 (2017) 29-36. DOI: 10.5815/ijisa.2017.09.04.