FacetRules: Discovering and Describing Related Groups
Lebna V. Thomas, Jiahao Deng, Eli T. Brown · 2021
Domain experts, owing to their knowledge and experience, develop an intuition for patterns in the data. They may know, for example, certain points of interest. However, they may not know exactly how to characterize what makes these data special. In our work, we start from these points of interest as seeds to derive groups of similar points automatically based on surrounding cluster structure. To aid characterization, we provide descriptive rules as an interpretable model to describe the groups. We explain this technique and present a prototype to demonstrate with a usage scenario how the technique helps the user with data exploration by discovering and describing groups in the data.