Багатофакторне прогнозування статистичних трендів для задач data science
The article deals with the processes of multi-factor forecasting of statistical trends for Data Science problems. Most of the classic approaches to data processing consist of studying the consequences of phenomena rather than the factors of their appearance. At the same time, the factors affecting t...
Збережено в:
| Дата: | 2024 |
|---|---|
| Автори: | , , , |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2024
|
| Теми: | |
| Онлайн доступ: | https://journal.iasa.kpi.ua/article/view/287739 |
| 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_ | 1866302930609504256 |
|---|---|
| author | Pysarchuk, Oleksii Andreieva, Tetiana Grinenko, Olena Baran, Danylo |
| author_facet | Pysarchuk, Oleksii Andreieva, Tetiana Grinenko, Olena Baran, Danylo |
| author_sort | Pysarchuk, Oleksii |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2024-08-11T01:12:49Z |
| description | The article deals with the processes of multi-factor forecasting of statistical trends for Data Science problems. Most of the classic approaches to data processing consist of studying the consequences of phenomena rather than the factors of their appearance. At the same time, the factors affecting the behavior of the investigated process are assumed to be random and are not investigated. The article discusses the approach to forecasting the parameters of the trend of statistical time series, which consists of the study of factors that lead to changes in the dynamics of the studied process. This approach potentially has better indicators of adequacy, accuracy, and efficiency in obtaining final solutions than classical approaches. The implementation of this approach is shown using an example of the analysis of exchange rate changes. The obtained results show the practicality of considering multi-factoriality in forecasting tasks. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2024.2.02 |
| first_indexed | 2025-07-17T10:28:20Z |
| format | Article |
| fulltext |
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran, 2024
Системні дослідження та інформаційні технології, 2024, № 2 21
UDC 004.5
DOI: 10.20535/SRIT.2308-8893.2024.2.02
MULTI-FACTOR FORECASTING OF STATISTICAL TRENDS
FOR DATA SCIENCE PROBLEMS
O. PYSARCHUK, T. ANDREIEVA, O. GRINENKO, D. BARAN
Abstract. The article deals with the processes of multi-factor forecasting of statisti-
cal trends for Data Science problems. Most of the classic approaches to data proc-
essing consist of studying the consequences of phenomena rather than the factors of
their appearance. At the same time, the factors affecting the behavior of the investi-
gated process are assumed to be random and are not investigated. The article dis-
cusses the approach to forecasting the parameters of the trend of statistical time se-
ries, which consists of the study of factors that lead to changes in the dynamics of
the studied process. This approach potentially has better indicators of adequacy, ac-
curacy, and efficiency in obtaining final solutions than classical approaches. The
implementation of this approach is shown using an example of the analysis of ex-
change rate changes. The obtained results show the practicality of considering multi-
factoriality in forecasting tasks.
Keywords: Data Science, multi-factor forecasting, statistical trends, currency rate
forecasting.
INTRODUCTION
The development of information technologies has led to their implementation in
many areas. One of the leading directions is the prediction of the indicators be-
havior of a certain controlled event. The examples of that can be: forecasting fluc-
tuations in currency markets; control of changes in economic performance indica-
tors of trading companies; forecasting the development of the epidemiological
situation; forecasting parameters of the technical state of equipment of production
lines, aviation systems, etc. All the listed applied tasks have the technological
unity of Data Science stages: data acquisition (measurement); their accumulation
(storage); data processing for the purpose of obtaining information about the
models and behavior of the researched process (processing, forecasting); extrac-
tion of knowledge and its manipulation [1; 2]. Currently, the focus of Data Sci-
ence issues is not on accumulation (measurement, storage), but on data processing
with the aim of extracting from them adequate, accurate and operational informa-
tion and knowledge. These processes in applied aspects of information technolo-
gies (IT) take place in the field of Big Data arrays and are manifested in the de-
velopment of Back-End components of distributed ERP / CRM software systems
with intellectual properties.
The key requirement of consumers for the final IT product is high quality in-
dicators of the source information, which are manifested in strict requirements for
the adequacy, accuracy and efficiency of the final solutions. It is possible to im-
plement this only in the direction of applying effective mathematical models for
processing Big Data arrays.
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 22
The experience shows that most classical approaches to data processing, re-
gardless of their classes, directions of improvement and effective implementation
to applied software systems, show their limitations [3–5]. They consist in the
study of the consequences of phenomena, and not the factors of their appearance.
For example, determining the trend and forecasting changes in the exchange rate
based on the results of a retrospective analysis of their behavior. At the same
time, the factors affecting the exchange rate are assumed to be random and are not
investigated.
Therefore, there is a need to implement R&D processes for the development
of mathematical support for modern ERP / СRM software systems capable of
meeting the high demands of consumers regarding the adequacy, accuracy and
efficiency of final solutions.
The article will consider an approach to predicting parameters of the trend of
statistical time series, which potentially has better indicators of adequacy, accu-
racy and efficiency of obtaining final solutions, compared to classical approaches.
Analysis of existing approaches. In its formulated form, we have the classic
task of applied statistical analysis / statistical learning: to build a mathematical
model based on a statistical sample of data, that ensures the determination of pre-
dictive values for the process being studied [1–5]. The key hypothesis in this is
the assumption of the random nature of the factors that affect the stochastic fluc-
tuations of each discrete dimension and, accordingly, determine the behavior of
the studied process outside the observation interval. As a rule, this happens due to
the complex and sometimes unknown nature of cause-and-effect relationships,
which determine the actual appearance of stochastic deviations and the develop-
ment of the situation in the future. Overcoming this a priori uncertainty is classi-
cally implemented through assumptions about the general appearance of the trend
model and the determination of its variables using complex algorithms, but the
principle hypothesis of randomness remains unchanged. That is, the primary sto-
chastic formalization of the problem has certain limitations in the accuracy of the
final result, which are determined by data processing methods.
Formulation of the problem. Therefore, the task of improving the methods
of statistical analysis / training in the direction of a detailed description and study
of factors that lead to the essence of the change in consequences – the dynamics
of the researched process – is urgent. The article examines the processes for mul-
tifactor forecasting of statistical trends for Data Science tasks. This is imple-
mented in the applied field of economic analysis of exchange rate changes. The
transition in statistical education from the analysis of consequences to factors re-
quires the implementation of a complex of R&D processes: the formation of an
informational model of factors that influence the change in currency rates; the
establishment of indicators (indicators describing change) of factors and criteria;
the measurement of indicators; and the statistical processing of indexes / indica-
tors (determination of statistical characteristics, construction of a trend line,
forecasting).
Thus, the goal of the article is the implementation of a complex of R&D
processes for multifactor forecasting of statistical trends for Data Science tasks
using the example of currency exchange analysis.
Multi-factor forecasting of statistical trends for Data Science problems
Системні дослідження та інформаційні технології, 2024, № 2 23
AN OVERVIEW OF THE MAIN MATERIAL
1. To form the infographic model of factors that influence on the change of
the currency exchange rates. The ratio of the dollar (USD) to the hryvnia (UAH)
was chosen as the exchange rate (hereinafter referred to as the exchange rate). On
the basis of the cognitive analysis of primary sources [6–13] and the practice of
currency trading, the factors affecting the exchange rate were determined.
Table 1. An infographic model of factors that influence the change in currency
rates
N Factor
group
N Factor
in the group
Indicator Data source, frequency
of measurement
The official exchange rate of the
hryvnia against the US dollar
Saldo of transactions of the natural
person on the sale/purchase
of foreign currency
The official website
of the NBU[10], daily
1
Sale/
purchase
of foreign
currency
1
Volume of
sale/purchase of
foreign currency
Saldo of NBU interventions The official website
of the NBU[10], weekly
Wheat export volume
Barley export volume
Rye export volume 1
Volume
of the main
Ukrainian
export goods Corn export volume
Website
of the Ministry
of Agrarian Policy
and Food
of Ukraine[8], daily
Wheat export price
Barley export price
2
Export prices for
the main
agricultural
products of
Ukraine
Corn export price
Fenix Agro
company website:
fenix-agro.com;
weekly
Hot-rolled steel export price
Armature export
Scrap steel export price
2
Export
of goods
3
Export prices
for the main
metal products
of Ukraine Iron ore raw materials export
Information and
analytical resource
about industry:
gmk.center, daily
Oil global price
3
Import
of goods 1
Global prices
for the main
imported goods Natural gas global price
The website
of the Ministry
of Finance [12], daily
The volume of hryvnia
government bonds in circulation
at nominal and amortized
cost with non-residents
The official website
of the NBU[10], daily
4
Foreign
investments 1
Participation of
non-residents in
trading in hryvnia
bonds of the
domestic state loan
The amount of funds involved
in the state budget for placement
of domestic government bonds
Website of the Ministry
of Finance
of Ukraine [9], weekly
Interest rates on deposits
in the national currency
1
The level of inter-
est rates on the
interbank market Interest rates on deposits
in US dollars
The official website
of the NBU[10], daily
NBU Key Policy Rate
5
Interest
rates
2 Interest rates
on deposits Ukrainian Overnight I
ndex Average (UONIA)
The website
of the Ministry
of Finance [12], daily
1 Stock indices
of Ukraine UX index
The website
of the Ministry
of Finance [12], daily 6
Stock
Market
2 World stock
indices Dollar index Investing.com, daily
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 24
2. To set the indicators (parameters which describes the change) of factors
and criteria is implemented as a result of the transformation of Table 1, based on
the essence of a specific factor.
T a b l e 2 . Indicators / parameters that describe the change of the factors and
criteria
№ Group of factors № Indicators Denotation Criterion
1
The official exchange rate of the
hryvnia
against the US dollar
ψ ψ→min
2
Saldo of transactions of the
natural person on the sale/ pur-
chase of foreign currency
φ φ→max
1
Sale/ purchase
of foreign
currency
3 Saldo of NBU interventions χ χ→min
4 Wheat export volume EVW EVW →max
5 Barley export volume EVB EVB→max
6 Rye export volume EVR EVR→max
7 Corn export volume EVC EVC→max
8 Wheat export price EPW EPW→max
9 Barley export price EPB EPB→max
10 Corn export price EPC EPC→max
11 Hot-rolled steel export price EPS EPS→max
12 Armature export EPA EPA→max
13 Scrap steel export price EPJ EPJ→max
2
Export
of goods
14 Iron ore raw materials export EP0 EP0→max
15 Oil global price IPOIL IPOIL→min
3
Import
of goods 16 Natural gas global price IPGAS IPGAS→min
17
The volume of hryvnia
government bonds in circulation
at nominal and amortized
cost with non-residents
INVV INVV→max
4
Foreign
investments
18
The amount of funds involved in
the state budget for placement of
domestic government bonds
INVM INVM→max
19
Interest rates on deposits in the
national currency
RDG RDG→max
20
Interest rates on deposits in US
dollars
RDD RDD→min
21 NBU Key Policy Rate P P→max
5 Interest rates
22
Ukrainian Overnight Index Aver-
age (UONIA)
UONIA UONIA→max
23 UX index UX UX→max
6 Stock Market
24 Dollar index DX DX→min
3. The indicators in Table 2 were measured on June 1, 2021. – November 1,
2022 according to the sources and frequency (discreteness) specified in Table 1.
The result is a multidimensional Big Data array of a statistical training sample of
24 indicators of 156 values, 5 (weekly monitoring) – of 36 values. Technological
Multi-factor forecasting of statistical trends for Data Science problems
Системні дослідження та інформаційні технології, 2024, № 2 25
efficiency of further processing processes is ensured by saving the received data
segment in the * format. xlsx file.
4. The statistical processing of indicators / parameters is implemented in the
sequence of classical stages of statistical training: determination of statistical
characteristics, construction of a trend line, forecasting. To increase the effective-
ness of statistical training, a hierarchy of interconnected alternative and innova-
tive stages is proposed (see the structural diagram in Figure). The structural
scheme takes into account the features of multi-factor forecasting of statistical
trends for Data Science tasks.
The data obtaining (block 1 of the diagram, Figure) is implemented quickly
from external sources using Web Scraping technologies.
Determination of the statistical characteristics of the obtained samples
(block 2) is carried out a posteriori in the format of calculation: expected value,
dispersion, standard deviation (SD), construction of a histogram of the law of dis-
tribution of the obtained samples. At the same time, the presence of a trend line is
taken into account, which is removed using the Least Square Method (LSM) with
a polynomial regression model [4].
Block 3 is intended for cleaning the statistical sample from anomalies. The
use of three algorithms for detecting and cleaning anomalies [15] increases the
reliability of the implementation of this process. Depending on the number of
anomalies, the strategy of rejecting them is used (up to 10% of anomalous meas-
urements – empirically obtained limits) and the recovery strategy (in other cases).
Optimizing the selection of the order of the trend line model (block 4) [14] is
implemented with the control of the values of three indicators, which also in-
creases the reliability of the final decisions.
Structural diagram of the multi-factor forecasting process of statistical trends for Data
Science tasks
Trend building
LSM polynomial
regression
LSM non-liner model
5
Forecasting
Polynomial regression
Non-linear model
6
Processes for 24 indicators
Integrated indicator
from 24 partial indicators
7
АM: sliding_wind
algorithm
АM: LSM
algorithm
АM: medium
algorithm
S
ol
ut
io
ns
2
o
f
3 Recovery
of AM
Rejection
of AM
3
Web Scraping:
statistics of the
* xlsx file
Finding statistical
characteristics
1
2
Analysis and processing
of abnormal measurements (AM)
Global
deviation
Control of
derivatives
Reliability of
approximation
S
ol
ut
io
ns
2
o
f
3
4
Optimization of the model
order
Processes for i-indicator
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 26
The global linear deviation of the estimate is one that compares across
multiple options:
.ˆ
1
1
1
n
i
ii yy
n
The accuracy of approximation 2R (coefficient of determination) varies
within 0...1 and should be minimal:
n
i
i
n
i
ii
yy
yy
R
1
2
1
2
2
)(
)ˆ(
1 ,
where n is a sample size;
n
i
ii y
n
y
1
1
, iy is a measured value; iŷ is LSM of
estimating the measured value.
The derivatives of the higher orders are the controls of obtaining small values:
,
)1()1(
1)(
t
yy
y
p
j
p
jp
j
,...1 mj np ...1 .
Determination of the trend line and forecasting (blocks 5, 6) is carried out
using the algorithm of the least squares method (LSM) in classical polynomial
[3; 4] or R&D nonlinear forms [4; 5].
For the presented research results, a nonlinear in parameters – transcendental
model was chosen
tbtactf sincos),( 00 ,
where },,{ 00 bac are the unknown parameters of the model. The procedure
for determining the parameters of a nonlinear model consistent with the measured
values is discussed in detail in [4; 5].
The calculation of the integrated assessment (unit 7) of the effect of factors
on the controlled parameter — the exchange rate is carried out according to the
scheme of multi-criteria / multi-factor assessment (SCOR) according to the
nonlinear scheme of compromises [16]. The data format is a multidimensional
discrete set of functions of 24 indicators.
According to the structural diagram of Fig. 1, an alpha version of the com-
puting unit (Backend component) of the ERP system layout was created to sup-
port currency trading processes. The software component is implemented in the
high-level python programming language with the use of technologies and librar-
ies: Web Scraping, pandas — for obtaining data; numpy — for “raw” program-
ming of data processing algorithms; matplotlib — for visualization of calculation
results.
Multi-factor forecasting of statistical trends for Data Science problems
Системні дослідження та інформаційні технології, 2024, № 2 27
THE RESULTS OF THE CALCULATIONS AND THEIR ANALYSIS
1. The official exchange rate of the hryvnia against the US dollar
2. Saldo of operations of physical persons on the sale/purchase of foreign
currency
The indicator is calculated as the difference between the sale of foreign cur-
rency and its purchase.
3. Saldo of NBU interventions
The indicator is calculated as the difference between the purchase of US
dollars and their sale.
The volume of the main agricultural products of Ukrainian exports (indica-
tors 4, 5, 6, 7) was calculated as the total volume of exported products, starting
from June 1, 2021 (the beginning of the study)).
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 28
4. Wheat export volume
5. Barley export volume
6. Rye export volume
7. Corn export volume
Multi-factor forecasting of statistical trends for Data Science problems
Системні дослідження та інформаційні технології, 2024, № 2 29
8. The export price of wheat
9. The export price of barley
10. The export price of corn
The export prices for all key agricultural products of Ukraine currently have
a positive trend.
11. Export price of hot rolled steel
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 30
12. The export price of armature
13. Export price of scrap metal
14. Export price for raw iron ore
15. The global oil price
Multi-factor forecasting of statistical trends for Data Science problems
Системні дослідження та інформаційні технології, 2024, № 2 31
16. The global natural gas price
17. The volume of hryvnia government bonds in circulation at nominal and
amortized cost with non-residents
18. The amount of funds involved in the state budget for placement of do-
mestic government bonds
19. The interest rates on deposits in the national currency
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 32
20. The interest rates on deposits in the USA dollars
21. NBU Key Policy Rate
22. Ukrainian Overnight Index Average (UONIA.
23. UX index
Multi-factor forecasting of statistical trends for Data Science problems
Системні дослідження та інформаційні технології, 2024, № 2 33
24. The dollar index
CONCLUSIONS
The real data obtained and processed allow us to identify useful features.
Statistical properties: parameters 1, 2, 3, 10, 11, 12, 23 (see histograms of distri-
bution laws) are characterized by a normal distribution law, the others have
combinatorial laws. This demonstrates the decomposition of the factors
influencing the exchange rate into unitary and combinatorial components. Inher-
ent natural presence of anomalous values of controlled parameters. The trend of
the studied indicators is non-linear, and the dynamics of change may be conflict-
ing according to the minimax analysis; that is, the improvement of certain indica-
tors may be accompanied by the deterioration of others. So, once the infological
model is formed, multifactorial consideration of the forecasting problem is appro-
priate. Further research will include the formation of an integrated indicator from
partial factors and a comparison of its dynamics with the dominant effect, the ex-
change rate. At the same time, one should expect an increase in the accuracy and
adequacy of predictive estimates of the studied parameters.
REFERENCES
1. Foster Provost and Tom Fawcett, Data Science for Business. Printed in the United
States of America. Published by O’Reilly Media, Inc., 2013, 409 p.
2. David Dietrich, Barry Heller, and Beibei Yang, Data Science & Big Data Analytics:
Discovering, Analyzing, Visualizing and Presenting. Indianapolis, Indiana: Data
John Wiley & Sons, Inc., 2015, 420 p.
3. Trevor Hastie, Robert Tibshirani, and Jerome Friedman, The Elements of Statistical
Learning Data Mining, Inference, and Prediction; 2nd edition. Springer, 2020, 768 p.
4. S.V. Kovbasyuk, O.O. Pisarchuk, and M.Yu. Rakushev, The least squares method
and its practical application. Zhytomyr: NAU, 2008, 228 p.
5. O.O. Pisarchuk, V.P. Kharchenko, Nonlinear and multicriterial modeling of
processes in traffic control systems. K.: Institute of Gifted Child, 2015, 248 p.
6. F.O. Zhuravka, Monetary policy in the context of transformational changes in
Ukraine’s economy: monograph. Sumy: “Business Perspectives,” Ukrainian
Academy of Banking of the National Bank of Ukraine, 2008, pp. 63–123.
7. S. Kulitsky, “Dynamics of US dollar exchange in Ukraine in 2019: attempted
situational analysis,” Ukraine: Events, Facts, Comments, no. 11, pp. 35–46, 2019.
Available: http://nbuviap.gov.ua/images/ukraine/2019/ukr11.pdf
8. Ministry of Agrarian Policy and Food of Ukraine [official site]. Available:
https://minagro.gov.ua
9. Ministry of Finance of Ukraine [official site]. Available: https://mof.gov.ua
10. National Bank of Ukraine [official site]. Available: http://bank.gov.ua
O. Pysarchuk, T. Andreieva, O. Grinenko, D. Baran
ISSN 1681–6048 System Research & Information Technologies, 2024, № 2 34
11. I. M. Sysoyeva, “The enterprise’s profit forecasting depending on accounting policy
methods,” Economics and State, no. 10, pp. 93–94, 2010. Available:
http://nbuv.gov.ua/UJRN/ecde_2010_10_26
12. Ukrainian Information Portal on Finance and Investments “Minfin.com.ua”. Avail-
able: https://minfin.com.ua
13. A. Khivrenko, Exchange rate: how it is determined, who influences it and what should be
guided. Available: https://www.epravda.com.ua/publications/2020/09/19/665288/
14. O.O. Pysarchuk, O.V. Korochkin, and D.R. Baran, “Determining the order of a
polynomial model for constructing a trend line in Data Science problems,” Problems
of Informatization and Management, 3(71), pp. 35–40, 2022. doi: 10.18372/2073-
4751.71.17001.
15. O. Pysarchuk, Yu. Mironov, I. Pysarchuk, and D. Baran, “Algorithms of Statistical
Anomalies Clearing for Data Science Applications,” System Research & Information
Technologies, no. 1, pp. 78–84, 2023. doi: 10.20535/SRIT.2308-8893.2023.1.06.
16. O. Pysarchuk, A. Gizun, A. Dudnik, V. Griga, T. Domkiv, and S. Gnatyuk, “Bifurca-
tion Prediction Method for the Emergence and Development Dynamics of Informa-
tion Conflicts in Cybernetic Space,” Proceedings of the International Workshop on
Cyber Hygiene (CybHyg-2019) co-located with 1st International Conference on Cy-
ber Hygiene and Conflict Management in Global Information Networks (CyberConf
2019). Kyiv, Ukraine, November 30, 2019, pp. 692–709. Available: https://ceur-
ws.org/Vol-2654/paper54.pdf
Received 04.09.2023
INFORMATION ON THE ARTICLE
Oleksii O. Pysarchuk, ORCID: 0000-0001-5271-0248, National Technical University
of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail: Plati-
numPA2212@gmail.com
Tetiana V. Andreieva, ORCID: 0009-0009-7033-9054, National Aviation University,
Ukraine, e-mail: tetyanaandreieva@gmail.com
Olena O. Grinenko, ORCID: 0000-0001-9673-6626, National Aviation University,
Ukraine, e-mail: gsa_ck@ukr.net
Danylo R. Baran, ORCID: 0000-0002-3251-8897, National Technical University of
Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Ukraine, e-mail:
danil.baran15@gmail.com
БАГАТОФАКТОРНЕ ПРОГНОЗУВАННЯ СТАТИСТИЧНИХ ТРЕНДІВ ДЛЯ
ЗАДАЧ DATA SCIENCE / О.О. Писарчук, Т.В. Андреєва, О.О. Гріненко, Д.Р. Баран
Анотація. Розглянуто процеси багатофакторного прогнозування статистичних
трендів для задач Data Science. Більшість класичних підходів до оброблення
даних полягають у дослідженні наслідків явищ, а не факторів їх появи. При
цьому фактори, що впливають на поведінку досліджуваного процесу, вважа-
ються випадковими та не досліджуються. Розглянуто підхід до прогнозування
параметрів тренду статистичних часових рядів, який полягає в дослідженні
факторів, що призводять до зміни динаміки досліджуваного процесу. Такий
підхід потенційно має кращі показники адекватності, точності і оперативності
отримання кінцевих рішень порівняно з класичними підходами. Наведено реа-
лізацію цього підходу на прикладі аналізу зміни курсу валют. Отримані ре-
зультати розрахунків показують доцільність розгляду багатофакторності
у задачах прогнозування.
Ключові слова: Data Science, багатофакторне прогнозування, статистичні
тренди, прогнозування курсу валют.
|
| id | journaliasakpiua-article-287739 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:28:20Z |
| publishDate | 2024 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/88/3ac70b232cc87602c55035dbcacb1788.pdf |
| spelling | journaliasakpiua-article-2877392024-08-11T01:12:49Z Multi-factor forecasting of statistical trends for data science problems Багатофакторне прогнозування статистичних трендів для задач data science Pysarchuk, Oleksii Andreieva, Tetiana Grinenko, Olena Baran, Danylo Data Science багатофакторне прогнозування статистичні тренди прогнозування курсу валют Data Science multi-factor forecasting statistical trends currency rate forecasting The article deals with the processes of multi-factor forecasting of statistical trends for Data Science problems. Most of the classic approaches to data processing consist of studying the consequences of phenomena rather than the factors of their appearance. At the same time, the factors affecting the behavior of the investigated process are assumed to be random and are not investigated. The article discusses the approach to forecasting the parameters of the trend of statistical time series, which consists of the study of factors that lead to changes in the dynamics of the studied process. This approach potentially has better indicators of adequacy, accuracy, and efficiency in obtaining final solutions than classical approaches. The implementation of this approach is shown using an example of the analysis of exchange rate changes. The obtained results show the practicality of considering multi-factoriality in forecasting tasks. Розглянуто процеси багатофакторного прогнозування статистичних трендів для задач Data Science. Більшість класичних підходів до оброблення даних полягають у дослідженні наслідків явищ, а не факторів їх появи. При цьому фактори, що впливають на поведінку досліджуваного процесу, вважаються випадковими та не досліджуються. Розглянуто підхід до прогнозування параметрів тренду статистичних часових рядів, який полягає в дослідженні факторів, що призводять до зміни динаміки досліджуваного процесу. Такий підхід потенційно має кращі показники адекватності, точності і оперативності отримання кінцевих рішень порівняно з класичними підходами. Наведено реалізацію цього підходу на прикладі аналізу зміни курсу валют. Отримані результати розрахунків показують доцільність розгляду багатофакторності у задачах прогнозування. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2024-06-28 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/287739 10.20535/SRIT.2308-8893.2024.2.02 System research and information technologies; No. 2 (2024); 21-34 Системные исследования и информационные технологии; № 2 (2024); 21-34 Системні дослідження та інформаційні технології; № 2 (2024); 21-34 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/287739/301034 |
| spellingShingle | Data Science багатофакторне прогнозування статистичні тренди прогнозування курсу валют Pysarchuk, Oleksii Andreieva, Tetiana Grinenko, Olena Baran, Danylo Багатофакторне прогнозування статистичних трендів для задач data science |
| title | Багатофакторне прогнозування статистичних трендів для задач data science |
| title_alt | Multi-factor forecasting of statistical trends for data science problems |
| title_full | Багатофакторне прогнозування статистичних трендів для задач data science |
| title_fullStr | Багатофакторне прогнозування статистичних трендів для задач data science |
| title_full_unstemmed | Багатофакторне прогнозування статистичних трендів для задач data science |
| title_short | Багатофакторне прогнозування статистичних трендів для задач data science |
| title_sort | багатофакторне прогнозування статистичних трендів для задач data science |
| topic | Data Science багатофакторне прогнозування статистичні тренди прогнозування курсу валют |
| topic_facet | Data Science багатофакторне прогнозування статистичні тренди прогнозування курсу валют Data Science multi-factor forecasting statistical trends currency rate forecasting |
| url | https://journal.iasa.kpi.ua/article/view/287739 |
| work_keys_str_mv | AT pysarchukoleksii multifactorforecastingofstatisticaltrendsfordatascienceproblems AT andreievatetiana multifactorforecastingofstatisticaltrendsfordatascienceproblems AT grinenkoolena multifactorforecastingofstatisticaltrendsfordatascienceproblems AT barandanylo multifactorforecastingofstatisticaltrendsfordatascienceproblems AT pysarchukoleksii bagatofaktorneprognozuvannâstatističnihtrendívdlâzadačdatascience AT andreievatetiana bagatofaktorneprognozuvannâstatističnihtrendívdlâzadačdatascience AT grinenkoolena bagatofaktorneprognozuvannâstatističnihtrendívdlâzadačdatascience AT barandanylo bagatofaktorneprognozuvannâstatističnihtrendívdlâzadačdatascience |