A Proposal for a Method to Systematically Integrate LLMs into Data Analytics Process
Lähde, Markus (2025)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025101826166
https://urn.fi/URN:NBN:fi:amk-2025101826166
Tiivistelmä
Large Language Models (LLMs) have been evaluated in many data analytics–related tasks, including complex table parsing, automatic statistical analysis, meaningful information recognition and tabular data modeling, synthesis and understanding. Several studies suggest enhancing these capabilities with the use of agentic architecture. However, a systematic method for integrat-ing LLMs into the data analytics process is currently lacking.
This thesis proposes such a method. In this study, a literature review was conducted to recognize core data analytics process steps. These steps were reformatted for concise process description with clear outputs for each step. Then a methodology for integrating LLM use into the data analytics process was crafted, expressed in a machine-readable format. The proposed method provides a foundation for future research on agentic frameworks and offers a tool for supporting cooperative human–LLM data analytics workflows.
This thesis proposes such a method. In this study, a literature review was conducted to recognize core data analytics process steps. These steps were reformatted for concise process description with clear outputs for each step. Then a methodology for integrating LLM use into the data analytics process was crafted, expressed in a machine-readable format. The proposed method provides a foundation for future research on agentic frameworks and offers a tool for supporting cooperative human–LLM data analytics workflows.