HAXplorer: Interactive visual exploration of hierarchical item and attribute spaces

Abstract

Analyzing tabular data by leveraging hierarchical structures for its items and attributes is a promising approach to scale for dataset sizes that make per-item and per-attribute analysis impractical. Existing approaches face limitations in supporting both item and attribute hierarchies, enabling user-controlled hierarchy creation, and ensuring visual scalability and interaction utility. We present HAXplorer, a visual analytics approach that enables users to create both item and attribute hierarchies, and to explore the resulting tabular data space by leveraging item and attribute aggregates. We demonstrate the generalizability of HAXplorer through usage scenarios across three diverse domains and evaluate its usefulness in a task-based user study. Usability is assessed through a perceived readability questionnaire and qualitative feedback. In addition to introducing a novel visual analytics system, our work offers insights into visual literacy, design validation methodologies, the positioning of HAXplorer within the broader landscape of biclustering techniques, and highlights the generative power of abstraction.