Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах
This article presents an algorithm for predicting the rate of a selected cryptocurrency, taking into account the posts of a group of famous people in a particular social network. The celebrities chosen as experts, i.e., famous personalities whose posts on social networks were studied, are either fam...
Збережено в:
| Дата: | 2023 |
|---|---|
| Автори: | , , |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2023
|
| Теми: | |
| Онлайн доступ: | https://journal.iasa.kpi.ua/article/view/285400 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Репозитарії
System research and information technologies| _version_ | 1866302916955996160 |
|---|---|
| author | Bidyuk, Petro Gavrilenko, Olena Myagkyi, Mykhailo |
| author_facet | Bidyuk, Petro Gavrilenko, Olena Myagkyi, Mykhailo |
| author_sort | Bidyuk, Petro |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2023-08-07T15:49:29Z |
| description | This article presents an algorithm for predicting the rate of a selected cryptocurrency, taking into account the posts of a group of famous people in a particular social network. The celebrities chosen as experts, i.e., famous personalities whose posts on social networks were studied, are either familiar with the financial industry, particularly the cryptocurrency market, or some cryptocurrency. The dataset used was the actual rates of the cryptocurrency in question for the selected period and the statistics of expert posts in the selected social network. The study used methods such as the full probability formula and the Bayesian formula. It was found that posts by famous people on social media differently affected cryptocurrency rates. The “main” expert was identified, and his posts were used to forecast the selected cryptocurrency’s rate. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2023.2.02 |
| first_indexed | 2025-07-17T10:28:16Z |
| format | Article |
| fulltext |
P. Bidyuk, O. Gavrilenko, M. Myagkyi, 2023
22 ISSN 1681–6048 System Research & Information Technologies, 2023, № 2
TIДC
ПРОГРЕСИВНІ ІНФОРМАЦІЙНІ ТЕХНОЛОГІЇ,
ВИСОКОПРОДУКТИВНІ КОМП’ЮТЕРНІ
СИСТЕМИ
UDC 681.513.7; 519.688
DOI: 10.20535/SRIT.2308-8893.2023.2.02
THE ALGORITHM FOR PREDICTING
THE CRYPTOCURRENCY RATE TAKING INTO ACCOUNT
THE INFLUENCE OF POSTS OF A GROUP OF FAMOUS
PEOPLE IN SOCIAL NETWORKS
P. BIDYUK, O. GAVRILENKO, M. MYAGKYI
Abstract. This article presents an algorithm for predicting the rate of a selected
cryptocurrency, taking into account the posts of a group of famous people in a par-
ticular social network. The celebrities chosen as experts, i.e., famous personalities
whose posts on social networks were studied, are either familiar with the financial
industry, particularly the cryptocurrency market, or some cryptocurrency. The data-
set used was the actual rates of the cryptocurrency in question for the selected period
and the statistics of expert posts in the selected social network. The study used
methods such as the full probability formula and the Bayesian formula. It was found
that posts by famous people on social media differently affected cryptocurrency
rates. The “main” expert was identified, and his posts were used to forecast the se-
lected cryptocurrency’s rate.
Keywords: cryptocurrency exchange rate, forecasting algorithms, social media
posts, group of experts, “main” expert, information technology of intelligent analysis.
INTRODUCTION
The study of cryptocurrency changes is gaining more and more popularity every
day due to the relative ease of entry and the abundance of recommendation infor-
mation regarding the process. Buying and selling cryptocurrencies is a rather in-
teresting process, since, if certain conditions are met, you can increase your
wealth several times, or even replace your main job with this business. However,
in order to really make money on this process, it is necessary to conduct research
on the chosen cryptocurrency, exchange and news related to them.
The relevance is due to the growing popularity of investing in cryptocurren-
cies. Publications of famous people who have a vested interest in this process
have a significant impact on the price formation of certain cryptocurrencies.
When traders create forecasts of changes in the exchange rate of certain crypto-
currencies, they will need a recommendation information system that can analyze
the impact of such publications on cryptocurrency changes, which will increase
the accuracy of the forecast.
The algorithm for predicting the cryptocurrency rate taking into account the influence …
Системні дослідження та інформаційні технології, 2023, № 2 23
The obtained forecasts can be used by financial market participants to obtain
high-quality forecasts of cryptocurrency rates, on the basis of which they will
make decisions on its purchase (sale).
LITERATURE ANALYSIS AND PROBLEM STATEMENT
The task of analyzing publications from the Internet is very important, since a
well analyzed publication can provide a lot of different information.
Article [1] discusses the process of computer based detection and categoriza-
tion of opinions expressed in a piece of text to determine whether the writer’s atti-
tude towards a particular topic, product, etc. is positive, negative, or neutral. A
detailed study of sentiment analysis and its cause-and-effect relationship. Using
sentiment analysis, you can get a generalized event based on mood and time. On
the other hand, the use of causality will be useful not only for determining the
causes and effects, respectively, but also for their further forecasting. The main
part of the article is an overview of the combination of these two approaches,
which degenerates into a model that allows to determine the mood for future
events, as well as to create a temporal forecast of the time that will pass between
certain events. To assess the accuracy, we used the following statistic: average
relative error.
To view publications, you need to choose a place where there are the most of
them and they are in a single text format, for this purpose Twitter is a good mes-
senger, work [2] discusses in detail the special linguistic analysis and statistics of
Twitter. This study aimed to identify criminal elements in the United States by
modeling topics of discussion and then incorporating them into a crime prediction
model. Thus, the study was conducted on the impact of social media posts on fu-
ture crimes.
In [3], methods for predicting user ratings of individual items using probabil-
istic algorithms were considered. In fact, the article perfectly illustrates the exis-
tence of computational patterns in terms of what exactly network users like under
certain circumstances. In other words, this study emphasizes the impact of prob-
abilistic algorithms in the field of recommender systems, and provides an over-
view of key methods that have been successfully applied. The considered algo-
rithms for object classification allow solving the problem of predicting user
evaluation of content and its categorization, as well as improving existing meth-
odologies for building information systems.
Article [4] is quite relevant today due to the difficult epidemiological situa-
tion in the world. It analyzed microblogs on Twitter and proposed several meth-
ods for identifying messages. It was determined that over ten weeks of more than
five hundred thousand reports, their best model achieved a correlation of 0.78
with CDC statistics.
It is also necessary to highlight Internet blogs, where many people express
their own opinions and visions of certain problems, etc. Therefore, in [5], a study
was conducted to identify hate groups. The proposed approach is semi-automatic
and consists of four modules, namely: blog spider, information retrieval, network
analysis, and visualization. The study was conducted on the Xanga blog site. The
results of the analysis were to identify some interesting demographic and topo-
logical characteristics in hate groups and to identify at least two large communi-
ties in addition to smaller ones. The proposed approach is also appropriate for
studying hate groups and other related communities on blogs.
P. Bidyuk, O. Gavrilenko, M. Myagkyi
ISSN 1681–6048 System Research & Information Technologies, 2023, № 2 24
For business, the process of analyzing large amounts of data and understand-
ing the needs of most people is very important, as it directly affects the com-
pany’s revenue. Article [6] provides a constructive consideration of the problem
of business accumulation of large amounts of data and the problems of their intel-
lectual processing. The author provides a clear definition and explanation of the
terms “data mining” and “data intelligence”. As a result, an objective conclusion
was made about the expediency of using data mining to increase the competitive-
ness of enterprises.
It should be noted that all of the above works describe the methods used in
our study, but do not provide the results of forecasting currency rates, including
cryptocurrencies. Accordingly, the factors that influence them were not studied.
In [7], the authors study the main macroeconomic indicators of influence on
the US dollar exchange rate in Ukraine: purchase/sale of cash currency,
purchase/sale of non-cash currency, balance of purchase/sale of cash and non-
cash currency, current year inflation, nominal and real GDP, purchase/sale by
bank customers, transactions between banks, gross and net international reserves,
unemployment rate, discount (interest) rate, balance of foreign exchange
interventions, and volume of transactions of nominal value. The main economic
components of exchange rate formation were identified using the principal
components method. Using the statistical models ARIMA, Exponential
Smoothing and SSA, the values of the selected factors of influence are predicted.
The values of exchange rates are forecasted using regression models built by Fast
Tree, Fast Forest, Fast Tree Tweedie and Gam algorithms, and the obtained
values are tested for accuracy. This article did not forecast cryptocurrency rates
specifically and did not study the impact of such a factor as publications in social
media.
Article [8] analyzes the methods, areas of application, and approaches to an-
alyzing publications and forecasting events based on the collected data, and also
gives the concept of the impact of publications on changes in the cryptocurrency
rate. The relevance of the topic is substantiated and the possibilities of appropriate
application of the results of the work are described. The main stages of working
with event forecasting data are identified, namely: data pre-processing, further
analysis and forecasting. This article did not investigate the level of influence of
celebrity publications on social media on the cryptocurrency rate. In addition, we
considered forecasts based on the posts of only one expert.
Within the framework of the studies cited in [7, 8], information systems
were created to implement the above tasks of data mining.
Studies have shown that celebrity posts do have a significant impact on cryp-
tocurrency rates. This can be easily verified using classical statistical analysis
tools, in particular, by analyzing the correlations between real and predicted cryp-
tocurrency rates. However, it should be noted that each famous personality –
hereinafter referred to as an expert – has a different level of awareness in the fi-
nancial sector, and is also involved in the process of forming cryptocurrency rates
in different ways (i.e. some experts are directly related to a particular cryptocur-
rency, and some are not), so the level of their influence on the forecasted rate will
be different. Therefore, it is advisable to study the level of influence of different
experts on the forecasted cryptocurrency rates in order to further rank them. This
study will improve the accuracy of cryptocurrency rate forecasts.
The following are recommended as ranking parameters:
The algorithm for predicting the cryptocurrency rate taking into account the influence …
Системні дослідження та інформаційні технології, 2023, № 2 25
1) the number of posts of a particular famous person in social networks for
the period under consideration;
2) the accuracy of the forecasts obtained for each expert in relation to the ac-
tual cryptocurrency rate;
3) deviations from the respective forecasts obtained without taking into ac-
count social media posts.
In this article, the number of posts for each of the pre-selected well-known
persons in social networks for the period under consideration is taken as such a
parameter.
PURPOSE AND OBJECTIVES OF THE STUDY
The purpose of the study is to develop an algorithm for predicting the cryptocur-
rency rate based on the posts of a group of famous people in social networks.
This will make it possible to increase the reliability of cryptocurrency rate
forecasting.
To achieve this goal, the following tasks were set:
create a list of experts and calculate the frequency of posts by each of
them in a particular social network;
to identify the expert whose posts will have the greatest impact on the rate
of the selected cryptocurrency in a selected period of time;
to obtain a forecast of the cryptocurrency rate taking into account the
posts of the “main” expert;
control the accuracy of the forecast.
Figure shows the process of calculating the rate forecast for the selected
cryptocurrency:
MATERIALS AND METHODS OF THE STUDY
The object of the study is the forecast of cryptocurrency rates.
The information required to analyze the level of influence of social media
posts on cryptocurrency rates is a list of experts whose level of influence will be
studied, the time interval of the study, the number of posts made by each of the
experts in question during the specified period of time, as well as the actual cryp-
tocurrency rates for the relevant period.
The process of calculating the cryptocurrency rate forecast
P. Bidyuk, O. Gavrilenko, M. Myagkyi
ISSN 1681–6048 System Research & Information Technologies, 2023, № 2 26
The use of mathematical tools based on the full probability and Bayesian
formulas allows us to use this information to determine the expert who is more
likely to make posts during the period under consideration. We will call this ex-
pert the “main” expert.
So, the following information is required as input: a list of experts, the num-
ber of posts by each expert, and the real rates of the selected cryptocurrency for
the period under consideration.
Experts were selected as well-
known personalities who are either
knowledgeable in the field of finance in
general and cryptocurrencies in particu-
lar, or whose activities are somehow
related to a particular cryptocurrency, or
not.
A fragment of the dataset is shown
in Table 1. Table 1 shows the real rates
of the selected cryptocurrencies, which
were taken from the website of the Bi-
nance crypto exchange [9].
Table 2 shows the number of posts
by the selected experts in 10 hours on
the social network.
The data generated in this way is
the input for this study. As part of the
study, it is necessary to:
create a list of experts and count
the number of their posts on social media;
determine the “main” expert us-
ing the Bayesian formula;
to obtain a forecast of the cryptocurrency rate, taking into account the
posts of the “main” expert, based on the approach described in [8].
to calculate the accuracy of MAPE forecasts.
The use of these methods guarantees reliable results when predicting crypto-
currency rates and studying the level of influence on them by posts of famous
people in social networks.
To conduct statistical analysis and obtain results based on these methods, the
corresponding software was developed.
ALGORITHM FOR TAKING INTO ACCOUNT THE LEVEL OF INFLUENCE
OF POSTS BY SEVERAL FAMOUS PEOPLE IN SOCIAL NETWORKS ON THE
CRYPTOCURRENCY RATE
Forming a list of experts and counting the frequency of posts by each of them
in a particular social network
Task statement. From the set of users of a social network, we select a subset
1 2( , , , )nA a a a of users who satisfy the following requirements:
1) the users are famous persons;
T a b l e 1 . Fragment of the input data
Hours Real courses
1 467
2 475
3 516
4 533
5 508
6 510
7 525
8 512
9 514
10 514
T a b l e 2 . Fragment of input data
(continued)
Expert
Number
of posts
Number
of posts related
to cryptocurrency
Expert 1 9 6
Expert 2 7 5
Expert 3 4 2
The algorithm for predicting the cryptocurrency rate taking into account the influence …
Системні дослідження та інформаційні технології, 2023, № 2 27
2) they are active users of the social networks and have a large number of
subscribers;
3) they`ll have different primary professional interests;
4) all pairs of users ia and ra , , 1,2, ,i r n , do not maintain communica-
tion in the network (they are not friends, do not respond to each other’s posts).
We call such users experts.
Suppose that over a certain period of time, experts have made m posts in a
social network, and k of them are related to a certain cryptocurrency. We consider
the context of the posts to be arbitrary. For the specified period of time, expert 1a
published m1 posts, of which 1k posts are related to a certain cryptocurrency, ex-
pert 2a published m2 posts, of which 2k posts are related to a certain cryptocur-
rency, … , expert na published mn posts, of which nk posts are related to a cer-
tain cryptocurrency:
mmmm n 21 ;
1 2 nk k k k ,
where nmmm ,,, 21 are the frequencies of expert posts; 1 2, , , nk k k are the fre-
quencies of expert posts related to a particular cryptocurrency, where i is the
number of the expert, 1,2, ,i n .
It is necessary to calculate the frequencies of posts of all selected experts for
an arbitrary time interval [10].
Rationale. This choice of experts is due to the need to form the set of such
experts who will be independent of each other both in the space of the chosen so-
cial network and in the professional space.
Results. From the data presented in Table 2, it can be seen that the consid-
ered set of 3 experts A = (expert 1, expert 2, expert 3). According to the social
networks data, it is known that 20m posts were published over a period of 10
hours, with 13k posts related to the selected cryptocurrency: ,91 m
4,7 32 mm , and 1 2 36, 5, 2k k k .
Determining the expert whose posts will have the greatest impact on the rate
of the selected cryptocurrency in a selected period of time
Task statement. Based on the list of experts 1 2( , , , )nA a a a obtained in sec-
tion 1 and taking into account the frequencies of their posts in the selected social
network for a specified small period of time — 1 2, , , nm m m , and 1 2, , , nk k k , it
is necessary to determine the “main” expert.
The formulated problem can be easily interpreted as a classical probabilistic
problem: m posts were written in a certain period of time. It is known that n ex-
perts published posts during this period, where 1 2, , , nm m m are the frequencies
of expert posts, nkkk ,,, 21 are the frequencies of expert posts related to the
chosen cryptocurrency, where i is the number of the expert, 1,2, ,i n . Action
A is that in an arbitrary period of time someone wrote the post related to the se-
P. Bidyuk, O. Gavrilenko, M. Myagkyi
ISSN 1681–6048 System Research & Information Technologies, 2023, № 2 28
lected cryptocurrency. It is necessary to determine which expert is more likely to
have made this post [11].
Rationale:
A (the post related to the selected cryptocurrency was written at any time t
from the interval [0;T]),
1H (the post was written by expert 1),
2H (the post was written by expert 2),
...
nH (the post was written by expert n).
We assume that actions iH are pairwise independent, where i is the number
of the expert, 1,2, ,i n . These assumptions can be made based on a list of re-
quirements that experts must meet (see section 1).
According to the full probability formula:
)/()((A)
1
ii
n
i
HAPHPP
, (1)
where
,)(
m
m
HP i
i (2)
where mi is the number of publications made by the expert i, and ik is the number
of publications made by the expert i related to the selected cryptocurrency, m is
the total number of publications for the period [0; ]T , )( iHP is the probability
that the post was published by expert i, )/( iHAP is the probability that at any
point in time t the post related to the selected cryptocurrency was written, pro-
vided that the post was written by expert i , 1,2, ,i n .
Then, using the Bayesian formula, we calculate the probability for each ex-
pert that he or she made the post, if it is known that the post was made during the
period under consideration:
)(
)/()(
)/(
AP
HAPHP
AHP ii
i , (3)
where )( iHP is the probability that the post was published by expert i, )/( iHAP
is the probability that at any point in time t the post related to the selected crypto-
currency was written, provided that the post was written by expert i, )/( AHP i is
the probability that the post was written by expert i, provided that it is known that
at any point in time t the post related to the selected cryptocurrency was written.
Among the obtained a posteriori probabilities, the highest one is chosen.
This means that this expert is most likely to have published a post in the time pe-
riod under consideration and thus will have a greater impact on the rate of the se-
lected cryptocurrency. It is this expert that will be considered the “main” expert
for the forecasted time period [ ; ]T T t . This is due to the fact that the influence
of the posts made during the time period [0; ]T also extends to a certain time pe-
riod [ ; ]T T t .
The algorithm for predicting the cryptocurrency rate taking into account the influence …
Системні дослідження та інформаційні технології, 2023, № 2 29
It should also be noted that the obtained a posteriori probabilities can be fur-
ther used to find average estimates of the effectiveness of predictive adaptive al-
gorithms for changing the cryptocurrency rate under the influence of the sequen-
tial appearance of individual or group posts of experts over time. The creation of
such algorithms and, as a result, the corresponding intelligent technologies is the
subject of further research by the authors.
Result. As mentioned in section 1, according to the data presented in Table 2,
we consider the set of 3 experts A (expert 1, expert 2, expert 3). According to
the social networks data, it is known that 20 posts were published during a pe-
riod of 10 hours, 13 posts related to the selected cryptocurrency, with
1 2 39, 7, 4m m m and 1 2 36, 5, 2k k k .
According to formula (2):
20
4
)( ,
20
7
)( ,
20
9
)( 321 HPHPHP ,
,
9
6
)/( 1 HAP .
4
2
)/( ,
7
5
)/( 32 HAPHAP
According to formula (1)
.
20
13
20
2
20
5
20
6
)P( A
According to formula (3):
,
13
6
)/( 1 AHP ,
13
2
)/(,
13
5
)/( 32 AHPAHP
so, with probability of
13
6
the post was most likely made by expert 1. Therefore,
he was considered the “main” expert for this period of time. The next expert is
expert 2 according to the probability of
13
5
, and the last one is expert 3 according
to the probability of
13
2
.
Obtaining a forecast of the cryptocurrency rate taking into account the posts
of the “main” expert
Task statement. Let the set of rates of the selected cryptocurrency for the time
period [0; ]T be known as the set }{ txX , [0; ]t T . You need to get the set of
forecasts of the cryptocurrency rates taking into account the “main” expert chosen
in section 1 for the period [ ; ]T T t as the set { }tY y , [ ; ]t T T t .
Rationale. To obtain forecasts, we will use the ATAPSN (algorithm for tak-
ing into account posts in social networks), taking into account the posts of the
“main” expert [8].
The idea of the algorithm is to calculate the coefficient of significance of the
posts of the “main” expert tc , at time t from the interval [ ; ]T T t , which is cal-
culated by the formula:
,δt tt chc (4)
where tch is the tone of the “main” expert’s post:
P. Bidyuk, O. Gavrilenko, M. Myagkyi
ISSN 1681–6048 System Research & Information Technologies, 2023, № 2 30
,s negativethe post iif ,1
,s neutralthe post iif, 0
,s positivethe post iif , 1
tch (5)
δt is the accuracy of the previous forecast,
δt t ty x , (6)
where ty is the predicted value of the cryptocurrency exchange rate obtained
without taking into account posts on social networks, tx is the actual value of the
cryptocurrency exchange rate, where t is the point in time.
After determining the coefficient tc from (4)–(6), a forecast of cryptocur-
rency exchange rate changes will be created based on the available data for the
period of time.
1 1t t ty y c .
Result. For the “main” expert identified in section 2, expert 1, a 10-hour
forecast of the selected cryptocurrency rate was obtained (see Table 3).
T a b l e 3 . Forecast of the selected cryptocurrency
rate for the expert 1
Hours Forecast rates
1 466.19
2 473.44
3 513.78
4 534.78
5 508.31
6 508.6
7 525.67
8 510.85
9 512.55
10 515.43
Control of the accuracy of the obtained forecast
Task statement. For each of the experts selected in section 1, based on the
ATAPSN algorithm (see section 3), we made a forecast of the rate of the selected
cryptocurrency and indicated the actual rate for the period of time under consid-
eration. The resulting forecasts, along with the actual cryptocurrency rates, are
provided in the input dataset (see Table 1).
Based on the input dataset, the following statistical samples jYX , ( 1,3)j
of size s each (where s is the number of forecasts made at the selected time point)
should be formed:
X — the set of real cryptocurrency rates;
1Y — the set of predicted cryptocurrency rates obtained using the ATAPSN,
taking into account expert 1 posts;
The algorithm for predicting the cryptocurrency rate taking into account the influence …
Системні дослідження та інформаційні технології, 2023, № 2 31
2Y — the set of predicted cryptocurrency rates obtained using the ATAPSN,
taking into account expert 2 posts;
3Y — the set of predicted cryptocurrency rates obtained using the ATAPSN,
taking into account expert 3 posts.
For each expert, it is necessary to calculate the accuracy of the MAPE forecast.
Rationale. The average relative sampling error is calculated by the formula:
100
x
yx
s
1
МАРЕ
1 l
jlls
l
j
.
where xl are sample items X , jly are sample items jY , (1,3)j , (1, )l s is the
volume of samples X and jY [12].
Using this measure of forecast accuracy will allow us to control the quality
of the dataset and rank experts in terms of the accuracy of forecasts obtained from
their posts (see Table 4).
T a b l e 4 . Level of model adequacy
МАРЕ, % Forecast accuracy
less than 10 High
10–20 Good
20–40 Satisfactory
40–50 Poor
more than 50 Unsatisfactory
Totally, these values are dependent on the purpose of the forecast. It is up to
the researcher to set the limits of the accuracy indicator that satisfy him or her.
In case of low accuracy of the forecast, it is recommended to make changes
to the significance factor to make the analysis of further changes more accurate
(see section 3).
Result. For samples 31 YY , we obtain the following values for the coeffi-
cients MAPEj (see Table 5).
T a b l e 5 . Values of thecoefficiennts MAPE
X 1Y 2Y 3Y
467 466.19 497.8707174 502.2597
475 473.44 502.1439974 463.9021
516 513.78 506.4172773 490.0742
533 534.78 510.6905573 515.4008
508 508.31 514.9638372 523.135
510 508.6 519.2371172 528.5218
525 525.67 523.5103971 518.3018
512 510.85 527.7836771 515.8382
514 512.55 532.056957 529.6971
514 515.43 536.330237 521.8438
MAPEj 0.23406855 3.27736587 3.1429464
P. Bidyuk, O. Gavrilenko, M. Myagkyi
ISSN 1681–6048 System Research & Information Technologies, 2023, № 2 32
Table 5 shows that sample 1Y has the lowest MAPE (0.23%), the next sam-
ple 3Y (3.14%), and the biggest error is in sample 2Y (3.28%). It should be noted
that the accuracy of all forecasts is high, according to Table 4, which indicates the
quality of the built forecasting model.
According to the results obtained, it can be stated that in terms of forecast
accuracy, expert 1 has the most significant posts, next expert 3, and finally expert 2.
The results are fully consistent with the fact that expert 1 was chosen as the
“main” expert, whose forecasts are the most significant.
CONCLUSIONS
In this article, we presented a modification of the ATAPSN algorithm [8], which
allows taking into account the posts of a group of pre-selected experts and form a
list of requirements for them.
This approach allows to increase the accuracy of forecasts of the selected
cryptocurrency rates, which has been confirmed statistically.
This approach calculates the a posteriori probabilities that a post related to
the selected cryptocurrency was written by a particular expert during the forecast-
ing interval. They were used to determine the “main” expert.
The obtained a posteriori probabilities can be further used to find average es-
timates of the efficiency of predictive adaptive algorithms for changing cryptocur-
rency rates under the influence of the sequential appearance of individual or
group posts by experts. The creation of such algorithms and, as a result, the corre-
sponding intelligent technologies is the subject of further research by the authors.
It should be noted that there may be different “main” experts at different
time intervals.
To use this approach, it is recommended to consider small time intervals,
each of which allows you to more accurately determine your “main” expert. This
increases the accuracy of forecasts of the selected cryptocurrency rates over the
entire time interval.
Using the post frequencies in social networks as a parameter for determining
the influence of experts allows us to apply the classical apparatus of probability
theory, which guarantees the correctness of the results obtained.
The disadvantages include the fact that the accuracy of the forecast may
be negatively affected by an unsuccessfully selected time interval for which
the forecast was made, since it is not known in advance how long an expert’s
post will be affecting the cryptocurrency rate. This indicates the need for the
constant monitoring of both cryptocurrency rates and expert posts on social
networks.
The proposed algorithm is an intermediate step towards the creation of a
multi-expert model for forecasting cryptocurrency rates.
The algorithm for predicting the cryptocurrency rate taking into account the influence …
Системні дослідження та інформаційні технології, 2023, № 2 33
REFERENCES
1. P. Preethi, V. Uma, and A. Kumar, “Temporal Sentiment Analysis and Causal Rules
Extraction from Tweets for Event Prediction,” Procedia Computer Science, no. 48,
pp. 84–89, 2015. Available: https://doi.org/10.1016/j.procs.2015.04.154
2. Matthew S. Gerber, “Predicting crime using Twitter and kernel density estimation,”
Decision Support Systems, no. 61, pp. 115–125, 2014. Available:
https://doi.org/10.1016/j.dss.2014.02.003
3. O.V. Gavrilenko and I.V. Kuptsova, “Research of tasks of estimation of objects in
computer social networks by means of reliable algorithms [Text],” materials of the
8th international scientific and technical conference “Modern directions of devel-
opment of information and communication technologies and means of manage-
ment”, Kharkiv, April 26–27, 2018, pp. 67–68.
4. A. Culotta and B. Huberman, “Towards detecting influenza epidemics by analyzing
Twitter messages,” Proceedings of the First Workshop on Social Media Analytics,
ACM, 2010, pp. 115–122.
5. M. Chau and J. Xu, “Mining communities and their relationships in blogs: A study
of online hate groups,” J. Hum.-Comput. Stud., 65(1), pp. 57–70, 2007.
6. O.V. Gavrilenko, “Review of methods of hybridization of intellectual information
processing,” Scientific Bulletin of Kherson State University, series “ECONOMIC
SCIENCES”, issue 8, part 3, p. 1, pp. 204–207, 2014.
7. O.V. Gavrylenko, K.D. Novakivska, and O.A. Shumeiko, “Selection of the most in-
fluential economic factors for forecasting the US dollar exchange rate,” NTU Bulle-
tin, no. 54, pp. 26–35, 2022. doi: 10.33744/2308-6645-2022-4-54-026-035.
8. O. Gavrylenko, M. Miahkyi, and Y. Zhurakovskyi, “The task of analyzing publica-
tions to build a forecast for changes in cryptocurrency rates,” Adaptive Automatic
Control Systems, vol. 2, no. 41, pp. 90–99, 2022. Available: https://doi.org/
10.20535/1560-8956.41.2022.271349
9. Binance cryptocurrency exchange. [Online]. Available: https://www.binance.
com/en/strategy/spot/grid/DOGEUSDT. Accessed on: 20 April 2023.
10. M.V. Kartashov, Probability, processes, statistics. Kyiv: VPC Kyiv University,
2007, 504 p.
11. V.P. Denysiuk, V.M. Bobkov, T.A. Pogrebetska, and V.K. Repeta, Higher mathematics.
P4. Probability theory and mathematical statistics. Kyiv: NAU-druk, 2009, 256 p.
12. Time series. [Online]. Available: https://kstat.pnu.edu.ua/wpcontent/uploads/sites/
63/2018 /04/%D0%A7%D0%B0%D1%81%D0%BE%D0%B2%D1%96-%D1%
80%D1%8F%D0%B4%D0%B8.pdf. Accessed on 20 April 2023.
Received 01.05.2023
INFORMATION ON THE ARTICLE
Petro I. Bidyuk, ORCID: 0000-0002-7421-3565, Educational and Research Institute for
Applied System Analysis of the National Technical University of Ukraine “Igor Sikorsky
Kyiv Polytechnic Institute”, Ukraine, e-mail: pbidyuke_00@ukr.net
Olena V. Gavrilenko, ORCID: 0000-0003-0413-6274, National Technical University
of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail: ge-
lena1980@gmail.com
P. Bidyuk, O. Gavrilenko, M. Myagkyi
ISSN 1681–6048 System Research & Information Technologies, 2023, № 2 34
Mykhailo Yu. Myagkyi, ORCID: 0000-0002-8038-8839, National Technical University
of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail: mishamyag-
kiy3@gmail.com
АЛГОРИТМ ПРОГНОЗУВАННЯ КУРСУ КРИПТОВАЛЮТИ
З УРАХУВАННЯМ ВПЛИВУ ДОПИСІВ ГРУПИ ВІДОМИХ ЛЮДЕЙ
В СОЦІАЛЬНИХ МЕРЕЖАХ / П.І. Бідюк, О.В. Гавриленко, М.Ю. Мягкий
Анотація. Наведено алгоритм прогнозування курсу обраної криптовалюти, з
урахуванням дописів групи відомих особистостей в конкретній соціальній ме-
режі. Експертами з-поміж них обирали ті, чиї дописи в соціальних мережах
досліджувалися, та які обізнані з галуззю фінансів, зокрема з ринком крипто-
валют, або так чи інакше з певною криптовалютою. Як датасет використано
реальні курси криптовалюти за обраний період часу та статистику дописів
експертів в обраній соціальній мережі. У межах дослідження застосовано такі
методи, як формула повної ймовірності та формула Баєсса. З’ясовано, що до-
писи відомих людей в соціальних мережах по-різному впливають на курси
криптовалют. Визначено «основного» експерта з урахуванням дописів якого
отримано прогноз курсу обраної криптовалюти.
Ключові слова: курс криптовалюти, алгоритми прогнозування, пости в соціаль-
них мережах, група експертів, «головний» експерт, інформаційна технологія
інтелектуального аналізу.
|
| id | journaliasakpiua-article-285400 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:28:16Z |
| publishDate | 2023 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/42/e4d44cce2a4cea02db79d68efe144442.pdf |
| spelling | journaliasakpiua-article-2854002023-08-07T15:49:29Z The algorithm for predicting the cryptocurrency rate taking into account the influence of posts of a group of famous people in social networks Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах Bidyuk, Petro Gavrilenko, Olena Myagkyi, Mykhailo курс криптовалюти алгоритми прогнозування пости в соціальних мережах група експертів інформаційна технологія інтелектуального аналізу cryptocurrency exchange rate forecasting algorithms social media posts group of experts “main” expert information technology of intelligent analysis This article presents an algorithm for predicting the rate of a selected cryptocurrency, taking into account the posts of a group of famous people in a particular social network. The celebrities chosen as experts, i.e., famous personalities whose posts on social networks were studied, are either familiar with the financial industry, particularly the cryptocurrency market, or some cryptocurrency. The dataset used was the actual rates of the cryptocurrency in question for the selected period and the statistics of expert posts in the selected social network. The study used methods such as the full probability formula and the Bayesian formula. It was found that posts by famous people on social media differently affected cryptocurrency rates. The “main” expert was identified, and his posts were used to forecast the selected cryptocurrency’s rate. Наведено алгоритм прогнозування курсу обраної криптовалюти, з урахуванням дописів групи відомих особистостей в конкретній соціальній мережі. Експертами з-поміж них обирали ті, чиї дописи в соціальних мережах досліджувалися, та які обізнані з галуззю фінансів, зокрема з ринком криптовалют, або так чи інакше з певною криптовалютою. Як датасет використано реальні курси криптовалюти за обраний період часу та статистику дописів експертів в обраній соціальній мережі. У межах дослідження застосовано такі методи, як формула повної ймовірності та формула Баєсса. З’ясовано, що дописи відомих людей в соціальних мережах по-різному впливають на курси криптовалют. Визначено "основного" експерта з урахуванням дописів якого отримано прогноз курсу обраної криптовалюти. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2023-06-30 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/285400 10.20535/SRIT.2308-8893.2023.2.02 System research and information technologies; No. 2 (2023); 22-34 Системные исследования и информационные технологии; № 2 (2023); 22-34 Системні дослідження та інформаційні технології; № 2 (2023); 22-34 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/285400/279502 |
| spellingShingle | курс криптовалюти алгоритми прогнозування пости в соціальних мережах група експертів інформаційна технологія інтелектуального аналізу Bidyuk, Petro Gavrilenko, Olena Myagkyi, Mykhailo Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| title | Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| title_alt | The algorithm for predicting the cryptocurrency rate taking into account the influence of posts of a group of famous people in social networks |
| title_full | Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| title_fullStr | Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| title_full_unstemmed | Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| title_short | Алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| title_sort | алгоритм прогнозування курсу криптовалюти з урахуванням впливу дописів групи відомих людей в соціальних мережах |
| topic | курс криптовалюти алгоритми прогнозування пости в соціальних мережах група експертів інформаційна технологія інтелектуального аналізу |
| topic_facet | курс криптовалюти алгоритми прогнозування пости в соціальних мережах група експертів інформаційна технологія інтелектуального аналізу cryptocurrency exchange rate forecasting algorithms social media posts group of experts “main” expert information technology of intelligent analysis |
| url | https://journal.iasa.kpi.ua/article/view/285400 |
| work_keys_str_mv | AT bidyukpetro thealgorithmforpredictingthecryptocurrencyratetakingintoaccounttheinfluenceofpostsofagroupoffamouspeopleinsocialnetworks AT gavrilenkoolena thealgorithmforpredictingthecryptocurrencyratetakingintoaccounttheinfluenceofpostsofagroupoffamouspeopleinsocialnetworks AT myagkyimykhailo thealgorithmforpredictingthecryptocurrencyratetakingintoaccounttheinfluenceofpostsofagroupoffamouspeopleinsocialnetworks AT bidyukpetro algoritmprognozuvannâkursukriptovalûtizurahuvannâmvplivudopisívgrupivídomihlûdejvsocíalʹnihmerežah AT gavrilenkoolena algoritmprognozuvannâkursukriptovalûtizurahuvannâmvplivudopisívgrupivídomihlûdejvsocíalʹnihmerežah AT myagkyimykhailo algoritmprognozuvannâkursukriptovalûtizurahuvannâmvplivudopisívgrupivídomihlûdejvsocíalʹnihmerežah AT bidyukpetro algorithmforpredictingthecryptocurrencyratetakingintoaccounttheinfluenceofpostsofagroupoffamouspeopleinsocialnetworks AT gavrilenkoolena algorithmforpredictingthecryptocurrencyratetakingintoaccounttheinfluenceofpostsofagroupoffamouspeopleinsocialnetworks AT myagkyimykhailo algorithmforpredictingthecryptocurrencyratetakingintoaccounttheinfluenceofpostsofagroupoffamouspeopleinsocialnetworks |