BUSINESS INFORMATICS Vol. 16 No.

2 – 2022

DOI: 10.17323/2587-814X.2022.2.21.35

The development of a model for

a personalized learning path using
machine learning methods*
Ekaterina S. Morozevich a
E-mail: katyamorozevich@mail.ru

Vladimir S. Korotkikh b
E-mail: vskor@bk.ru

Yevgeniya A. Kuznetsova a
E-mail: zhenya.kuz-1997@yandex.ru

a
Reshetnev Siberian State University of Science and Technology
Address: 31, Krasnoyarsky Rabochy Av., Krasnoyarsk 660037, Russia
b
Municipal budget general education institution secondary school No. 4
Address: 9, st. Naberezhnaya, Divnogorsk 663091, Russia

Abstract
Today, the economy is undergoing a digital transformation. Its key barriers are a lack of
qualified personnel, competencies and knowledge, as well as internal resistance in organizations.
It can be overcome through quality staff development and training. An urgent problem is to build a
personalized learning path. Modern research is aimed at the implementation of recommendation
systems in order to select relevant material. However, these recommendations are based on digital
traces; the student’s full personal profile, as well as organizational values are not considered. This
study aims is to create an intelligent guide that would accompany an employee throughout his life
in the organization, involving him in the learning process according to a personalized path based
on a complex personal profile and reactions to educational material, training soft and hard skills
in accordance with the values of the organization and the employee. Methods of system analysis,
system engineering, psychodiagnostic research (the DISC model, Rowe’s “Decision-making Style”
methodology, Honey and Mumford’s method of determining activity styles, psychotype test),
software design and artificial intelligence (matrix factorization and neural networks) were used in

*
The article is published with the support of the HSE University Partnership Programme

21
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

this study. The study was conducted on a unique database collected as part of its implementation and
consisting of educational tasks for soft skills development, plus data on their implementation by users
with different soft skills profiles. An intelligent guide model has been developed and implemented as
a software component for an enterprise management system. The basis consists of psychodiagnostic
modules, organizational management, training and recommendations. The intelligence of the
system we developed allows you to qualitatively form a personalized learning path that will involve
an employee not only in the learning and development process, but also in achieving organizational
goals. The organization receives T-shaped specialists who have a proactive position and are capable
of self-organization by investing in the development of employees. The results of this study can
be used by enterprises not only at the organizational level, but also through broadcasting in the
education system to form an education ecosystem in accordance with the requirements of innovative
development of a given region’s economy.

Keywords: training and development, supra-professional competencies, adaptability,

recommendation system, machine learning, matrix factorization, neural network
Citation: Morozevich E.S., Korotkikh V.S., Kuznetsova Y.A. (2022) The development of a model
for a personalized learning path using machine learning methods. Business Informatics, vol. 16,
no. 2, pp. 21–35. DOI: 10.17323/2587-814X.2022.2.21.35

Introduction does not realize what is happening, because

he or she does not know what happened, and

T
oday, the economy is undergoing a does not guess what will happen. In view of
digital transformation. This is a rev- this, the urgent task is the development of per-
olutionary process of transforming sonnel, taking into account the points of con-
an organization’s business model, not only by tact between the values of the employee and the
using digital technologies, but also by intro- organization.
ducing fundamental organizational changes in
technology, culture, operations and the princi- Activities’ digitalization generates an increase
ples of creating new products. in data volume that requires further processing.
It is data that underlies digital transformation,
The connecting link between the integration
being the lifeblood of the transition process to
of deep organizational transformations and
the digital economy.
new technologies are people – employees with
their knowledge, skills and values. According Already in 2020, according to Data Age
to Deloitte Consulting LLC, 53% of organi- Report, about 51 zettabytes of information
zations already understand that in the current were generated by mankind, and by 2025 this
conditions from 50 to 100% of their employees data volume will increase almost 3.5 times and
should acquire new skills and abilities [1]. amount to 175 zettabytes [2]. This shows the
trend of digital data exponential growth.
Organizations are created by people with
their value system’s investment in the structure The users’ activity recorded by various
being created. An employee who comes to an devices represents their digital traces. Today,
organization at a certain stage of the life cycle digital footprints occupy a significant part of

22
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

the big data cloud. Their key direction is to researchers have begun to actively study using
extract information about the preferences of personal characteristics in educational recom-
potential customers and offer products which mendations. For example, the use of informa-
they will be interested in, that is, increasing tion about the learning style and emotional
sales [3]. reactions [13]. Moreover, the issue of the rec-
ommendation systems used in online learning
However, the information extracted from
is increasingly rising in the research [14].
digital traces can be used not only in external
interactions, but also in the internal activities of Thus, this study’s purpose is to create an
the company. Modern companies are increas- intelligent guide which would accompany an
ingly using data to increase employee engage- employee throughout his life in the organi-
ment in achieving and separating goals, that is, zation, involving him in the learning process
to increase the productivity of both employ- according to a personalized path based on data
ees and organizational processes, thereby cre- on the personal profile and reactions to educa-
ating new sources for their competitive advan- tional material, training soft and hard skills in
tage [4]. accordance with the values of the organization
and the employee.
Thus, recommendation systems are an inte-
gral part of the major market players today. The
recommendation system is a complex of algo- 1. An intelligent system
rithms, programs and services designed to form of personnel training
relevant recommendations to users regarding and development concept
the object of information search [5].
The principles of system engineering pro-
Many studies note [6–8] that it is possible vided the basis for the intelligent guide devel-
to overcome the key barriers of digital trans- opment for the employees’ growth and train-
formation in the form of a shortage of quali- ing. The requirements’ collection of interested
fied personnel, competencies and knowledge, parties to the system was being developed as the
as well as internal resistance in organizations breakaway point in the creation of an intelli-
[9–10] by creating a system of training and gent guide.
staff development. This system is based on the
As the result of analyzing the requirements,
formation of a personalized learning path [11]
it was revealed that the modular architecture of
relying on the values of the organization and
the system will not only take into account all
the employee and relevant educational mate-
the stakeholders’ requirements, but also differ-
rial, in other words, on those objects, actions,
entiate access to various functions depending
tasks, which the employee will perform with
on the positions held. The modular architec-
increased interest and, as a result, with maxi-
ture proposed for implementation in accord-
mum efficiency.
ance with the identified requirements is shown
In the education field, the use of information in Fig. 1.
systems based on recommendations is designed
A database was created using the MySQL
to solve one of the student’s main problems –
database management system to implement
choosing educational material. Often such
the intelligent guide and support the work of
material is training programs and educational
the proposed modules.
courses. When selecting educational material
in such systems, the student’s personal char- The diagnostic module’s main task is the
acteristics are usually not taken into account, formation of a digital profile of an employee,
but his preferences are used [12]. Recently, which will make it possible to take into account

23
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

educational task
Diagnostic module Training module Recommendations module

tasks execution history

employee’s profile Database employee’s profile, organization profile,

tasks execution history

Organizational management
module organizational profile

Fig. 1. Intelligent guide model.

all the characteristics of the individual and will izational structure is created through this
become the basis for effective training and module. After that, the head of the structural
development. unit determines the points of the critical and
desired profile of the employee for each posi-
Modern trends are such that organizations
tion under consideration, in accordance with
are increasingly focusing on supra-professional
the labor functionality and labor actions, as
skills or so-called soft skills, which are much
well as the supra-professional skills designated
more difficult to reproduce with the help of
on the basis of organizational values.
digital transformation tools and which under-
lie the ongoing changes [15]. Soft skills are the Thus, having compiled a position profile for
basis for the effective development of profes- each element of the organizational structure,
sional skills or so-called hard skills. a comprehensive profile of the organization
is formed which reflects all the features of its
Thus, it was proposed to form a digital profile
activities.
of an employee, consisting not only of personal
data (gender, age, education, position, work The educational process organization is the
experience), but also revealing the levels of basis of the training module. Users in the frame-
development of employee soft skills in demand work of soft skills development perform unique
in the labor market in the digital economy. educational tasks (Fig. 3). For each task, the
user evaluates how much he liked the task, its
Comprehensive automated diagnostics were
effectiveness and complexity. Thus, a data set
developed to determine the level of proficiency
is formed which characterizes the educational
in one or another soft skill (Fig. 2). The DISC
tasks’ performance by users and is used in the
model, Rowe’s “Decision-making Style”
future to form relevant recommendations.
methodology, Honey and Mumford’s method
of determining activity styles, and a psychotype The recommendations module is designed to
test were used as diagnostic material. These form a personalized learning path in accord-
methods reveal the psychoemotional charac- ance with the values of the employee and the
teristics of a person in 16 projections with their organization and to select the most relevant
respective competencies. educational material for the employee.
The organizational management module is It is possible to determine the soft skills
designed to fix organizational values through directions’ development of an employee within
formation of job profiles. First of all, an organ- the interests of the organization if there is an

24
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

Fig. 2. Diagnostic module.

employee and position profile. However, as The next step is the direct selection of rele-
mentioned earlier, it is necessary to synchro- vant educational material using a recommen-
nize the values of the employee and the organ- dation system. To select an algorithm for its
ization. So it is also necessary to determine operation, a study was conducted on technolo-
the desired development directions for the gies such as machine learning based on matrix
employee. Within the framework, an employee factorization and a neural network.
development needs questionnaire depending
on his value orientations has been developed. 2. Computational experiment
After the diagnostics, employee is asked to 2.1. Problem Statement
answer a list of questions to form the desired
employee profile and determine the compe- There are many users U = {u1, u2, ... , un} and
tencies that he would like to develop many educational tasks T = {t1, t2, ... , tm}.

The overlapping of the obtained profiles on The V matrix contains the scores assigned
each other makes it possible to visualize the by users to educational tasks. There will be a
level of competencies’ expression and to iden- number in place vij (i 1, ... n, j 1 ... m) if the ui
tify the directions of an employee’s personal user has evaluated the task tj and is empty oth-
development in accordance with the needs of erwise.
the organization, all of which leads to the for- It is required to find a vector containing the
mation of their common development path. user’s already known estimates vij , as well as the

25
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

Fig.3. Training module.

expected ones . Based on the obtained vector, ♦♦ matrix, which contains both latent user
you need to rank the list of educational tasks T characteristics and explicit ones (gender,
for the ui user. education, work experience and user profile
values consisting of 16 parameters);
2.2. Machine learning ♦♦ matrix, which is characterized educa-
based on matrix factorization tional tasks.
Let us fill in the latent characteristics of the W
Matrix factorization implies decomposition
and H matrices with random variables based on
of the original matrix into a product of two
the law of uniform distribution on the interval
matrices of small rank. The user’s interaction
with the object is modeled as a scalar product .
of the vectors of user’s imagine and the object Next, we will solve the minimization problem:
in the factor space. Factorization models work
effectively with highly sparse matrices [16–18]. (1)

Let us imagine the evaluation matrix V as a where is the matrix obtained by approxima-
product of two matrices: tion from W and H;

26
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

, – algorithm parameters. calculation time increases. Conversely, with a

value of k < 40, too rough an approximation is
Each algorithm’s iteration for finding a solu-
obtained.
tion to minimize the error consists of the fol-
lowing steps: Based on the test results (Table 1), with an
increase in the learning rate from 0.005 to 0.03,
1. Fix the H matrix.
there is an improvement in the results (a decrease
2. Find the error , j 1, ..., n. in error, an increase in prediction accuracy). At
3. Find new values , the value = 0.04, the results are usually equiva-
where p 1, ..., k; lent to the results at = 0.03 or worse. With the
– regularization parameter; number of iterations 20 ( = 0.03), high accu-
racy of factorization is already observed on the
– learning rate.
given samples. Based on these results, we use
4. Fix the W matrix. the following parameters for the algorithm:
5. Find the error . = 0.03; the number of iterations is 20.

6. Find new values , With an increase in the amount of data

where p 1, ..., k; received and a decrease in the sparsity of data,
it may be necessary to increase the number of
– regularization parameter;
iterations. The control of the number of itera-
– learning rate.
tions was automated by testing after the end of
When solving the problem experimentally, factorization and comparing the accuracy of the
the rank of the matrices W and H k = 40 was prediction with the previous testing. As the pre-
found. This value is a boundary value for a diction accuracy decreases, the number of itera-
given case in which the root-mean-square tions increases. Otherwise the number of itera-
and absolute error of solving the factorization tions remains the same. Another alternative is to
problem are optimal. With a value of k > 40, set the permissible accuracy of factorization, at
the advantage is not traced as a result, and the which the algorithm should be stopped.

Table 1.
Speed and factorization errors (n = 14000)

Iterations’ number 0.005 0.01 0.02 0.03 0.04

absolute average 0.197 0.119 0.064 0.05 0.048

20 the quadratic mean 0.24 0.09 0.013 0.007 0.0064

time, sec. 6 6 5 5 5
absolute average 0.113 0.057 0.028 0.02 0.022

40 the quadratic mean 0.082 0.013 0.002 0.001 0.001

time, sec. 11 10 10 11 11
absolute average 0.075 0.03 0.017 0.014 0.014

60 the quadratic mean 0.03 0.003 0.0009 0.0006 0.0006

time, sec. 16 15 16 15 15

27
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

One of the main tasks in the construction of the study, tasks for which there are no factors
recommendation systems is to solve the prob- are randomly offered to users.
lem of ‘cold’ start [19]. It occurs when new ele-
To assess the quality of predictions, the set of
ments appear in the system: whether it is a user
V estimates was divided into samples for Vtrain
or preference objects.
training and Vtest testing. The model was tested
In the problem under consideration, two var- with the following quantitative characteristics:
iants of cold start of users can be distinguished:
♦♦Users’ number: 303;
the user did not perform the educational tasks;
the user performed the educational tasks, but ♦♦Task’s number: 11726;
the factorization task was solved before his reg- ♦♦Task’s number completed by users: 17733;
istration; that is to say, there are no user factors
in the system. ♦♦Training sample: 75%;

Let us consider the first case. Nothing is ♦♦Test sample: 25%;

known about the user’s preferences. We know ♦♦Matrices’ rank: 40;
only about his profile. It is some vector which
♦♦Factorization steps: 20;
is formed from the values of gender, seniority,
position. If the user has passed the test, then ♦♦Range of original matrix’ values: 0–10 (nor-
the vector will also have a soft skills profile of malized data).
16 components. The V matrix is very sparse; most of the cells
The problem of cold start in this case is pro- are not filled (more than 95% of the cells).
posed to be solved using cosine similarity. To When evaluating the performance of the
do this, there is a cosine between the new user model, three types of approximation were car-
and each existing user. It is ranked in descend- ried out: for efficiency, complexity and prefer-
ing order, and the user to whom the maximum ences. Four cases were considered:
cosine corresponds is selected. As a result,
Using only latent factors (marked as FREE
recommendations are formed for a new user
on graphs);
already based on the similarity of the user
found through the cosine: Using fixed user factors along with latent
(soft skills profile, gender, position, education;
designated as UV);
(2)
Using fixed task factors along with latent ones
(the average values of performance indicators,
In the second case, the problem is proposed
complexity and preferences set by users, the
to be decided by solving the factorization prob-
level of complexity, the orientation of tasks to
lem for a specific user. In this case, the W and V
competencies, are designated as TV);
matrices will take the form of row vectors. The
problem factors H matrix will be fixed, and the Using fixed factors, users and tasks (des-
search for values will be performed only for ignated as UV+TV).
the W vector-string. The search algorithm is Fifty factorizations were carried out on ran-
reduced to a special case: i = 1, and the steps of dom data for each case. Then these data were
the matrix factorization algorithm from the 4th averaged. The testing also took into account
to the 6th are excluded. the cold start of users based on cosine similar-
We also need to note the ‘cold’ start problem ity. The cold start of tasks was not taken into
for educational tasks. At the current stage of account in testing.

28
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

per cent of sample per cent of sample

100 100

90 FREE 90 FREE
UV UV
80 80
TV TV
70 UV+TV 70 UV+TV

60 60

50 50

40 40

30 30

20 20

10 10

0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 1

Fig. 4. Error in predicting preference using Fig. 5. Error in predicting preference using
factorization on percentage intervals. factorization on discrete values [0; 1].

per cent of sample per cent of sample

100 100

90 FREE 90 FREE
80
UV UV
80
TV TV
70 UV+TV 70 UV+TV
60 60

50 50

40 40

30 30

20 20

10 10

0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 1

Fig. 6. Error in predicting preference using Fig. 7. Error in predicting preference using
cosine similarity on percentage intervals. cosine similarity on discrete quantities [0; 1].

The data for the preference prediction error The fixed factors’ use mainly worsens
are presented in Fig. 4–7. The data for the the result, as can be seen from the graphs
efficiency prediction error are presented in (Fig. 4–11). This may be due to the small size
Fig. 8–11. of the data set.

29
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

per cent of sample per cent of sample

100

90 FREE 70 FREE
UV UV
80 60
TV TV
70 UV+TV UV+TV
50
60

40
50

40 30

30
20
20
10
10

0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 1 2 3 4

Fig. 8. Error in predicting efficiency using Fig. 9. Error in predicting efficiency using
factorization on percentage intervals. factorization on discrete quantities [1; 5].

per cent of sample per cent of sample

40 70
FREE FREE
35 UV UV
60
TV TV
30 UV+TV UV+TV
50
25
40
29

30
15

20
10

5 10

0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 1 2 3 4

Fig. 10. Error in predicting efficiency using Fig. 11. Error in predicting efficiency using
cosine similarity on percentage intervals. cosine similarity on percentage intervals
and on discrete quantities [1; 5].

2.3. Neural network solving a wide range of tasks in various fields of

activity [21].
A neural network is a computational struc-
ture assembled into computational elements’ To solve this problem, such neural network
network created in the likeness of a biological architectures as a convolutional neural net-
neuron [20]. Neural networks are capable of work [22] and a multilayer perceptron accord-

30
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

ing to Rumelhart [23] were considered. The est accuracy: two hidden layers (the first layer –
latter proved to be the most effective. In addi- 64 neurons, the second layer – 128 neurons),
tion, a convolutional neural network requires the size of the batch – 32, the number of
more data and training time. epochs – 20.
Two methods were investigated: encoding The neural network results and the predic-
words with a unique number and embedding tion accuracy are presented in Table 2.
words (Word2Vec) to vectorize the educational To construct a vector, the network is launched
task [24]. Encoding words with a unique num- as many times as the values are contained in
ber is faster, but it does not allow you to record the database of educational tasks. Thus, after
any relationships between words, as well as the all iterations, a vector is formed on the basis
similarity of words. of which it is possible to select the most pre-
Using the Word2Vec model for word embed- ferred tasks for the user.
ding allows you to capture subtle relationships
between words, but requires a large amount of 3. A comparison
data for training and is a lengthy process. Due of Artificial Intelligence
to the absence of the need for constant vec- technologies
torization of educational tasks, but only when
replenishing the database of educational tasks, Two approaches to the formation of recom-
the Word2Vec model was used. mendations were considered: neural network
and machine learning based on matrix factori-
The input layer of the neural network has
zation. Both approaches have both advantages
one neuron, which receives a vector containing
and disadvantages.
data about the user and tasks.
As opposed to vectors fed to the neural
The activation function ‘ReLU’ is activated
network input, the factor matrix consists of
on the hidden layers, the activation func-
unique user vectors that contain latent features
tion ‘Sigmoid’ is activated on the output layer,
that are not taken into account by the neural
which allows you to amplify weak signals and is
network. The neural network, in turn, relies
not saturated by strong signals [25].
only on a given known user vector consisting
During training and testing the neural net- of parameters laid down by the researcher. It
work, a data set consisting of 16  862 vectors allows us to talk about a more personalized
containing data about users and tasks was used. approach when using matrix factorization.
The data were divided into training and test
The neural network also requires the manda-
samples in the proportion of 80/20.
tory availability of not only all data about the
On the studied data set, the neural network user, but also their correctness. When using
with the following topology showed the great- machine learning based on matrix factoriza-

Table 2.
Neural network results

Preference Complexity Effectiveness

Standard deviation 0.056 0.102 0.098

Accuracy, % 93.31 68.79 61.0

31
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

tion, you can make recommendations after tion has been carried out in the form of a soft-
several evaluations without having any more ware component for the enterprise manage-
information. ment system.
The matrix factorization’s disadvantage is a The new competencies’ development will
cold start, when the user has not yet evaluated lead to the expansion of the employee’s pro-
any tasks, and he needs to give recommen- file and ability to perform new tasks in the area
dations. However, this problem is solved very of interest, thereby obtaining a high-quality
quickly using the method described earlier. result. The organization, in turn, will receive
new opportunities for development in the
Despite the fact that the researchers were market and rapid adaptation to changing con-
faced with the task of ranking, it was necessary ditions. Thus, by investing in employee devel-
to compare the work of these methods accord- opment, the organization receives T-shaped
ing to three criteria: preferences (binary value), specialists with a proactive position and capa-
complexity (discrete [1; 4]) and efficiency (dis- ble of self-organization. This directly leads to
crete [1; 5]). The comparison results are pre- the realization of the potentials of both the
sented in Table 3. employee and the organization.
As can be seen from Table 3, on discrete val- The intelligent guide has a modular archi-
ues, the accuracy of neural network predictions tecture, which is due to the interdiscipli-
decreases compared to matrix factorization. narity and novelty of the research, as well as
Based on the above, the researchers decided the simplicity of implementing the received
to use machine learning based on matrix fac- developments into real management prac-
torization as the core of the recommendation tice. This approach made it possible to con-
module. duct qualitative research in two projections:
the development of comprehensive auto-
Conclusion mated soft skills diagnostics of an employee’s
profile and relevant filling of an individual
The problem of low adaptability of manage- educational path with educational material
ment in the conditions of digital transforma- using artificial intelligence, and then to com-
tion was considered in this article. Within the bine the results into a single system for train-
research framework, a model of an employ- ing and staff development.
ee’s intelligent guide has been developed. This In the proposed approach artificial intelli-
model allows us to level this problem due to gence is expressed by machine learning using
the qualitative development of the employ- matrix factorization. Such intelligence allows
ee’s competence profile, taking into account you to qualitatively select educational material
the value orientations of the employee and which will be interesting to the employee from
the organization. The model’s implementa- his personal positions and maximally involve

Table 3.
Accuracy of artificial intelligence technologies’ predictions

Preference Complexity Effectiveness

Neural network 93.3% 68.8 61%

Matrix factorization 93.1% 76.4 62.4%

32
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

him not only in the learning and development considering downtime not as a loss, but as an
process, but also in achieving organizational opportunity for the development and training
goals. of employees.
The proposed database structure allows you The developed software component can be
to collect digital traces of users describing bio- used by enterprises not only at the organiza-
graphical characteristics, skills and level of tional level, but also through broadcasting in
their proficiency at a certain stage of develop- the education system. On the one hand, enter-
ment, the history of choosing and completing prises will be able to reduce staff downtime,
educational tasks, and further use of the col- expand the personal competence profile of
lected data as the basis for the work of the rec- employees, and increase the growth of the pro-
ommendations module and the formation of duction potential of both employees and enter-
personalized proposals. prises in the market. On the other hand, edu-
The interdisciplinarity and novelty of the cational institutions will be able to train highly
research also determine the variability of its qualified personnel within the framework of
development. As part of further research, it is orientation to the values of the market and the
planned to integrate the software component values of students with the formation of indi-
into the enterprise’s business process man- vidual educational trajectories. Such coopera-
agement system. This is supposed to predict tion will make it possible to form an education
the time when the labor resource is not loaded ecosystem in accordance with the require-
and offer educational tasks to fill it. It will be ments of innovative development of a region’s
the basis of a fundamentally new approach to economy.

References
1. Deloitte Insights (2020) Results of the study “International trends in human resource management –
2020” in Russia. Available at: https://www2.deloitte.com/content/dam/Deloitte/ru/Documents/
human-capital/russian/hc-trends-2020-Russia.pdf (accessed 22 November 2021) (in Russian).
2. TAdviser (2020) Big Data global market. Available at: https://tadviser.com/a/e.php?id=129607
(accessed 22 November 2021).
3. Fatyanov A.A. (2018) Big Data in the digital economy: its value and legal challenges. Economics. Law.
Society, no. 4, pp. 37–40 (in Russian).
4. Reinsel D., Gantz J., Rydning J. (2018) The digitization of the world – from edge to core. IDC White
Paper. Available at: https://www.seagate.com/files/www-content/our-story/trends/files/idc-seagate-
dataage-whitepaper.pdf (accessed 22 November 2021).
5. Ali N.M., Gadallah A.M., Hefny H.A., Novikov B.A. (2021) Online web navigation assistant. Vestnik
Udmurtskogo Universiteta. Matematika. Mekhanika. Komp’uternye Nauki, vol. 31, no. 1, pp. 116–131.
https://doi.org/10.35634/vm210109
6. Gokhberg L. (ed.) (2019) What is the digital economy? Trends, competencies, measurement. Moscow:
HSE (in Russian).
7. Kulikova M.Kh., Kulikova M.Kh., Magomadov M.A. (2020) IT in education: today, tomorrow and always.
Proceedings of the I Student Scientific and Practical Conference “Information technologies in business and
education”, Grozniy, 20 February 2020, pp. 85–90 (in Russian). https://doi.org/10.36684/21-2020-1-32-35
8. Geissbauer R., Lübben E., Schrauf S., Pillsbury S. (2018) Digital Champions. How industry leaders
build integrated operations ecosystems to deliver end-to-end customer solutions. PwC Strategy& Global
Digital Operations Study. Available at: https://www.strategyand.pwc.com/gx/en/insights/industry4-0/
global-digital-operations-study-digital-champions.pdf (accessed 22 November 2021).

33
BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

9. Harvard Business Review (2017) High-performance sourcing and procurement driving value through
collaboration. Harvard Business School Publishing. Available at: https://hbr.org/resources/pdfs/
comm/scoutrfp/HighPerformanceSourcing.pdf (accessed 22 November 2021).
10. Dolganova O.I., Deeva E.A. (2019) Company readiness for digital transformations: problems and diagnosis.
Business Informatics, vol. 13, no 2, pp. 59–72. http://doi.org/10.17323/1998-0663.2019.2.59.72
11. Komarov V.A., Sarafanov A.V. (2021) IoT systems in the process of multidisciplinary training of
personnel for the digital economy and their design. Business Informatics, vol. 15, no 2, pp. 47–59.
http://doi.org/10.17323/2587-814X.2021.2.47.59
12. Rivera A.C., Tapia-Leon M., Lujan-Mora S. (2018) Recommendation systems in education:
A systematic mapping study. Proceedings of the International Conference on Information Technology
& Systems (ICITS 2018), Libertad City, Ecuador, 10–12 January 2018 (eds. A. Rocha, T. Guarda).
Advances in Intelligent Systems and Computing, vol. 721, pp. 937–947.
https://doi.org/10.1007/978-3-319-73450-7_89
13. Bustos López M., Alor-Hernández G., Sánchez-Cervantes J., Paredes-Valverde M., Salas-Zárate
M.P. (2020) EduRecomSys: An educational resource recommender system based on collaborative
filtering and emotion detection. Interacting with Computers, vol. 32, no. 4, pp. 407–432.
https://doi.org/10.1093/iwc/iwab001
14. Urdaneta-Ponte M.C., Mendez-Zorrilla A., Oleagordia-Ruiz I. (2021) Recommendation
systems for education: Systematic review. Electronics, vol. 10, no. 14, article ID 1611.
https://doi.org/10.3390/electronics10141611
15. Bughin J., Hazan E., Lund S., Dahlström P., Wiesinger A., Subramaniam A. (2018) Skill shift:
automation and the future of the workforce. McKinsey & Company. Available at: https://www.mckinsey.
com/featured-insights/future-of-work/skill-shift-automation-and-the-future-of-the-workforce
(accessed 22 November 2021).
16. Rubtsov V.N. (2017) Matrix factorization based on deep learning for collaborative filtering. Student
Theses. Moscow: HSE. Available at: https://www.hse.ru/en/edu/vkr/206744221 (accessed 22 November
2021) (in Russian).
17. Strömqvist Z. (2018) Matrix factorization in recommender systems: How sensitive are matrix factorization
models to sparsity. Uppsala University Publications. Available at: https://uu.diva-portal.org/smash/
get/diva2:1214390/FULLTEXT01.pdf (accessed 22 November 2021).
18. Mojsyuk-Dran’ko P.A., Revotyuk M.P. (2020) Matrix factorization methods for recommendation
systems. Proceedings of the International Conference Information Teсhnologies and Systems 2020 (ITS
2020). Minsk: Belarusian State University of Informatics and Radioelectronics, pp. 193–194. Available
at: https://libeldoc.bsuir.by/bitstream/123456789/41339/1/Moysyuk_Dranko_Metody.pdf (accessed
22 November 2021) (in Russian).
19. Kuznetsov I.A. (2019) Machine learning methods and algorithms for preprocessing and classification of
semi-structured text data in scientific recommendation systems. Moscow: NRNU MEPhI. Available at:
https://ds.mephi.ru/documents/90/Кузнецов_И_А_Текст_диссертации.pdf (accessed 22 November
2021) (in Russian).
20. Golovko V.A., Krasnoproshin V.V. (2017) Neural network data processing technologies. Minsk:
Belarusian State University. Available at: https://elib.bsu.by/bitstream/123456789/193558/1/Golovko.
pdf (accessed 22 November 2021) (in Russian).
21. Lisovsky A.L. (2020) Application of neural network technologies for management development of
systems. Strategic decisions and risk management, vol. 11, no. 4, pp. 378–389 (in Russian).
https://doi.org/10.17747/2618-947X-923
22. Bezdan T., Bacanin Džakula N. (2019) Convolutional neural network layers and architectures.
Proceedings of the Sinteza 2019: International Scientific Conference on Information Technology and Data
Related Research, Belgrade, 20 April 2019 (ed. Milovan Stanišić), pp. 445–451.
https://doi.org/10.15308/Sinteza-2019-445-451

34
 BUSINESS INFORMATICS Vol. 16 No. 2 – 2022

23. Postarnak, D.V. (2012) The critical analysis of models of neural networks. Vestnik Tyumenskogo
gosudarstvennogo universiteta, no. 4, pp. 162–167. Available at: https://elibrary.ru/item.
asp?id=17758787& (accessed 22 November 2021) (in Russian).
24. Young T., Hazarika D., Poria S., Cambria E. (2018) Recent trends in deep learning based natural
language processing. IEEE Computational Intelligence Magazine, vol. 13, no. 3, pp. 55–75.
https://doi.org/10.1109/MCI.2018.2840738
25. Iqbal T., Qureshi S. (2020) The survey: Text generation models in deep learning. Journal of King Saud
University – Computer and Information Sciences. https://doi.org/10.1016/j.jksuci.2020.04.001

About the authors

Ekaterina S. Morozevich
Process architect, Research Management Department, Reshetnev Siberian State University of Science
and Technology, 31, Krasnoyarsky Rabochy Av., Krasnoyarsk 660037, Russia;
E-mail: katyamorozevich@mail.ru
ORCID: 0000-0001-8796-3157

Vladimir S. Korotkikh
System administrator, Municipal budget general education institution secondary school No. 4, 9,
st. Naberezhnaya, Divnogorsk 663091, Russia;
E-mail: vskor@bk.ru
ORCID: 0000-0001-9270-3376

Yevgeniya A. Kuznetsova
Engineer, Research Management Department, Reshetnev Siberian State University of Science
and Technology, 31, Krasnoyarsky Rabochy Av., Krasnoyarsk 660037, Russia;
E-mail: zhenya.kuz-1997@yandex.ru
ORCID: 0000-0002-3559-8394

35