Abstract:
In the contemporary era, the surge in data-driven decision-making underscores the significance of extracting actionable insights from intricate data sources. Large Langua...Show MoreMetadata
Abstract:
In the contemporary era, the surge in data-driven decision-making underscores the significance of extracting actionable insights from intricate data sources. Large Language Models (LLMs) emerge as transformative entities reshaping data analytics paradigms. These models not only facilitate data insight generation but also enhance data visualization when integrated with specialized tools. With their extensive parameters, LLMs possess an innate ability to discern nuanced intricacies and patterns within data, transcending conventional analytic methods. Nonetheless, challenges arise, including the potential for generating non-factual information and navigating complex multi-step procedures. We introduces an approach to enhance LLMs analytic capacities. The main contribution is an agent-based graph where multiple agents collaboratively address data analytics challenges. These agents encompass data sources, Python libraries for data visualization, and even other LLMs for data digestion and summarization. Notably, this methodology enable LLMs as inquisitive agents, actively seeking clarity and engaging other agents for task completion. This dynamic, graph search, inter-agent communication framework promises a more comprehensive data analytics process with reduced human feedback.
Published in: 2023 International Conference on Artificial Intelligence, Blockchain, Cloud Computing, and Data Analytics (ICoABCD)
Date of Conference: 13-15 November 2023
Date Added to IEEE Xplore: 16 January 2024
ISBN Information: