Large Language Models for Transforming Categorical Data to Interpretable Feature Vectors.

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• Analyzing heterogeneous data with numerical and categorical attributes often requires separate treatment or transformation of data types.
• Transforming categorical attributes to numerical ones enables integrated multivariate analysis.

Purpose of the Study:

• To propose a novel technique for transforming categorical data into interpretable numerical feature vectors using Large Language Models (LLMs).
• To facilitate integrated multivariate analysis of heterogeneous datasets.
• To enable intuitive user adjustment and improve AI-generated outputs through an interactive tool.

Main Methods:

• Utilizing Large Language Models (LLMs) to identify key characteristics of categorical attributes.
• Assigning numerical values to these characteristics to generate multi-dimensional feature vectors.
• Developing an interactive tool for validating and refining AI-generated transformations.
• Proposing novel methods for ordering and color-coding categories based on feature vector similarities.

Main Results:

• Successful transformation of categorical data into interpretable numerical feature vectors using LLMs.
• Demonstration of a fully automated transformation process with options for intuitive user adjustment.
• Validation of the approach through an interactive tool.
• Development of new methods for category ordering and color-coding based on generated feature vectors.

Conclusions:

• LLMs offer a powerful and interpretable approach to transforming categorical data for integrated analysis.
• The proposed technique enhances data analysis by providing numerical representations of categorical attributes.
• The interactive tool empowers users to refine and validate AI-driven data transformations.
• Novel visualization methods can be derived from the generated feature vectors for improved data exploration.