Institutional Repository of Vilnius University
Investigation of process automation with large language models /
Abstract
dc:descriptionThis thesis investigates the role of Large Language Models (LLMs), specifically AutoGPT, in the automation of software development for data analysis, focusing on the impact of integrating functional and non-functional requirements on LLM decision quality. This study evaluates AutoGPT's ability to write Python scripts for data processing and to create R Shiny dashboards across three distinct datasets, using prompts of varying complexity. The experiment progresses from simple natural language prompts to structured ones, finally to prompts that integrate both functional and non-functional requirements. This progression enables an in-depth assessment of AutoGPT’s efficiency, accuracy and challenges. Key findings show that AutoGPT can autonomously generate Python scripts and R Shiny dashboards. The study reveals that prompt complexity enhances AutoGPT’s output quality and efficiency, although challenges such as the reproducibility of generated codes and sensitivity to the structure of prompts are observed. Concluding, the thesis underscores the potential and current limitations of AutoGPT in software development for data analysis. It suggests a strategic approach to prompt construction and indicates that while AutoGPT is promising, it does not yet match the capability of a skilled human data analyst.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Sokolovas, Manvydas,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192058310&prefLang=en_US