.png){kind=icon}
.png){kind=icon}
Return
Enhancing decision-making with large language models: context and sensitivity analysis
.png){kind=icon}
PRE
AI
DOI：10.1007/s12065-025-01113-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The speed of innovation of technologies around the world makes life easier, especially technologies that use the Internet (IoT). IoT-generated data often suffers from low quality due to issues such as duplicated data, missing values, and outliers, that negatively impact the analysis result. Data preprocessing is the approach that assists in enhancing the data quality in real time by tackling the issues. Setting the context in a more efficient way helps to improve the data cleaning process. This paper explores integrating LLMs into an interactive user interface for data stream cleaning, enhancing decision-making through context-aware recommendations and sensitivity assessments. The LLM operates locally, assisting users in dynamically refining contexts and sensitivity levels based on real-time interactions. To validate our proposed method, we conducted experiments using a healthcare dataset, evaluating results with precision, recall, and F1 score metrics, alongside the BERTScore for comparison. To ensure reliability and alignment with embedded instructions, we also engaged five expert reviewers. Beyond healthcare, we tested the proposed method on datasets from weather, business, industry, and education sectors, demonstrating its broad applicability. A comparative analysis with existing methods highlights the LLM's effectiveness in guiding users, ultimately fostering improved decision-making across diverse applications.
Keywords:
Context awareness recommendation
Data sensitivity level
Large language model (LLM)
Near real-time data stream
Journal
IF:
2.6
Papers:
116
Citations:
2.0K