УКР ENG

Search:


Email:  
Password:  

 REGISTRATION CERTIFICATE

KV #19905-9705 PR dated 02.04.2013.

 FOUNDERS

RESEARCH CENTRE FOR INDUSTRIAL DEVELOPMENT PROBLEMS of NAS (KHARKIV, UKRAINE)

According to the decision No. 802 of the National Council of Television and Radio Broadcasting of Ukraine dated 14.03.2024, is registered as a subject in the field of print media.
ID R30-03156

 PUBLISHER

Liburkina L. M.

 SITE SECTIONS

Main page

Editorial staff

Editorial policy

Annotated catalogue (2011)

Annotated catalogue (2012)

Annotated catalogue (2013)

Annotated catalogue (2014)

Annotated catalogue (2015)

Annotated catalogue (2016)

Annotated catalogue (2017)

Annotated catalogue (2018)

Annotated catalogue (2019)

Annotated catalogue (2020)

Annotated catalogue (2021)

Annotated catalogue (2022)

Annotated catalogue (2023)

Annotated catalogue (2024)

Annotated catalogue (2025)

Thematic sections of the journal

Proceedings of scientific conferences


Large Language Models: A New Paradigm for Data Analysis
Tumanov O. O.

Tumanov, Oleksii O. (2025) “Large Language Models: A New Paradigm for Data Analysis.” Business Inform 9:106–113.
https://doi.org/10.32983/2222-4459-2025-9-106-113

Section: Economic statistics

Article is written in Ukrainian
Downloads/views: 0

Download article (pdf) -

UDC 004.8:519.23

Abstract:
The emergence of large language models (LLMs) has fundamentally transformed the paradigm of data analysis, affecting fields such as social media analytics, healthcare, and software development. These models, examples of which include architectures like GPT and BERT, handle tasks related to unstructured text and contextual understanding very effectively, but their widespread adoption requires a critical assessment of their relationship to and differences from traditional statistical approaches. This article aims to provide a comparative analysis of LLMs and traditional statistical methods, examining their strengths, limitations, and areas of application. The study uses a qualitative comparison method based on the following criteria: data type (unstructured vs. structured), interpretability, purpose, data requirements, and reproducibility. The analysis shows that, despite the effectiveness of LLMs in natural language text processing, their probabilistic nature imposes significant limitations on reliability and ethics. Traditional statistical methods, in contrast, offer high interpretability and reproducibility, which are critically important for deriving deterministic conclusions. Special attention is paid to the limitations of both approaches. Specifically, LLMs are prone to «hallucinations» and inheriting biases from training data, which can lead to discriminatory outcomes. Meanwhile, traditional statistical models are not immune to selection biases. As a result, the study found that LLMs and traditional methods are not competitors but complementary tools, each best suited for addressing certain tasks. The main conclusion is that the most promising direction is the integration of these approaches into hybrid methodologies. Large language models (LLMs) can be effectively used for preprocessing and converting unstructured text into structured quantitative data, which can then be analyzed using classical statistical methods. This approach allows overcoming existing limitations and achieving greater reliability, transparency, and accuracy in scientific research.

Keywords: large language models, statistical methods, data analysis, artificial intelligence, ethical considerations, hybrid model.

Fig.: 5. Tabl.: 2. Bibl.: 12.

Tumanov Oleksii O. – PhD, Senior Lecturer, V. N. Karazin Kharkiv National University (4 Svobody Square, Kharkіv, 61022, Ukraine)
Email: [email protected]

List of references in article

CADE. (2025, August 14). EU report finds promise and hurdles in using AI for healthcare. https://cadeproject.org/updates/eu-report-finds-promise-and-hurdles-in-using-ai-for-healthcare/
Eurostat. (2024). Use of artificial intelligence in enterprises. https://ec.europa.eu/eurostat/statistics-explained/index.php/Use_of_artificial_intelligence_in_enterprises
GAO. (2025, July 29). Artificial Intelligence: Generative AI Use and Management at Federal Agencies (GAO-25-107653). https://www.gao.gov/products/gao-25-107653
GitHub Blog. (2021). Introducing GitHub Copilot: your AI pair programmer. https://github.blog/news-insights/product-news/introducing-github-copilot-ai-pair-programmer/
Ji, W., Yuan, W., Getzen, E., et al. (2025, February 25). An Overview of Large Language Models for Statisticians. arXiv. https://arxiv.org/html/2502.17814v1
Kumar, N., Ramavath, Sh. K., & Venkatesh, V. (2023). Bridging Context, Statistics, and Practice: A Multi-Dimensional Framework for Responsible LLM Evaluation and Selection. TechRxiv. https://doi.org/10.36227/techrxiv.175691248.85284358/v1
Moiseienko, M. I., Kuzyshyn, M. M., Turovska, L. V., & in. (2024). Velyki movni modeli shtuchnoho intelektu v medytsyni [Large language models of artificial intelligence in medicine]. Suchasni informatsiini tekhnolohii ta innovatsiini metodyky navchannia v pidhotovtsi fakhivtsiv: metodolohiia, teoriia, dosvid, problemy, 72, 73–88. https://doi.org/10.31652/2412-1142-2024-72-73-88
Potiatynyk, B. (2025). Vykorystannia ShI v analizi mediapotokiv: mozhlyvosti ta obmezhennia [The use of AI in the analysis of media flows: opportunities and limitations]. Visnyk Lvivskoho universytetu. Seriia «Zhurnalistyka», 57, 62–74. https://doi.org/10.30970/vjo.2025.57.13290
Tabnine. (2025). About | Tabnine: The AI code assistant that you control. https://www.tabnine.com/about/
Tumanov, O. O. (2020). Statystychni metody analizu danykh sotsialnykh media [Statistical methods of social media data analysis]. Biznes Inform, (2), 266–272. https://doi.org/10.32983/2222-4459-2020-2-266-272
Tumanov, O. O. (2021). Statystychne otsiniuvannia rozvytku sotsialnykh media [Statistical assessment of social media development] [Avtoreferat dysertatsii kandydata ekonomichnykh nauk, Natsionalna akademiia statystyky, obliku ta audytu]. http://nasoa.edu.ua/wp-content/uploads/zah/tumanov_avt.pdf
Tumanov, O. O. (2025). Rozrobka systemy statystychnykh pokaznykiv dlia monitorynhu emotsiinoho stanu u sotsialnykh media [Development of a system of statistical indicators for monitoring the emotional state in social media]. Materialy III Mizhnarodnoi naukovo-praktychnoi konferentsii «Skhidnoievropeiskyi tsentr naukovykh doslidzhen», 114–118. Research Europe.

 FOR AUTHORS

License Contract

Conditions of Publication

Article Requirements

Regulations on Peer-Reviewing

Publication Contract

Current Issue

Frequently asked questions

 INFORMATION

The Plan of Scientific Conferences


 OUR PARTNERS


Journal «The Problems of Economy»

  © Business Inform, 1992 - 2025 The site and its metadata are licensed under CC BY-SA. Write to webmaster