Lists
Lists
Patel, Preeti and Shooshtarian, Sowgol (2024) The role of chatbots in data analytics: an evaluation of functional abilities. In: 12th Computing Conference, 11th-12th July 2024, London. (Submitted)
In the fast-paced evolution of artificial intelligence, in which chatbots are becoming vital tools in many areas, this study is focused on the various functional roles of ChatGPT-4 in data analytics. The primary objective is to assess the technological advancements and capabilities of ChatGPT-4 in analyzing data. The research methodology employs three distinct datasets and the analysis compares various prompt engineering techniques with ChatGPT-4 against human-driven analysis. This comparison is grounded in profile optimization techniques, the CRISP-DM framework, and a structured Data Science Project Plan. The findings reveal the diverse capabilities of ChatGPT-4. On one hand, basic prompts yield quick, initial insights, positioning ChatGPT-4 as a valuable asset for those with limited technical expertise. Conversely, refined and engineered prompts enable ChatGPT-4 to deliver a deeper, more detailed analysis. Yet, when contrasted with traditional human-driven analysis, typically utilizing tried and tested techniques, tools and experience, all combined with a deep knowledge of machine learning and data analysis; the human approach remains unsurpassed in depth, granularity, and accuracy. Conclusively, while ChatGPT-4 stands out for preliminary data explorations and general insights, complicated, nuanced analytical tasks are best tackled using time-tested, traditional data science methodologies and tools. While ChatGPT-4 is versatile, it has not yet matched the expertise of seasoned data analysts.
![[img]](/style/images/fileicons/text.png)
Restricted to Repository staff only until 12 February 2025.
Download (3MB)
 |
View Item |