Improving Clustering Explainability and Automated Cluster Naming with Approximate Inverse Model Explanations and Large Language Models

This study presents a method that combines an explainable AI (XAI) technology called approximate inverse model explanation (AIME) with large language models (LLMs) to improve the explainability of clustering results and automatically generate cluster names. AIME identifies the features that are most important in forming each cluster. A language model then uses these important features to automatically generate names for the clusters. Traditionally, clustering has been used to group similar data; however, determining why clusters form and name them has been left to data analysis. This study automates both tasks by explaining cluster formation and generating cluster names, making it a useful tool in fields such as data mining. For example, in marketing, customer data can be clustered to identify segments, and the characteristics and names of these segments can be automatically generated. This method enhances the explainability of unsupervised learning and offers a new application for explainable AI. Although the method has been demonstrated for data mining, it has potential applications in other fields, such as text mining.