Visual Analysis of High-Dimensional Event Sequence Data via Dynamic Hierarchical Aggregation.

Area of Science:

• Computer Science
• Data Visualization
• Human-Computer Interaction

Background:

• Visual analytics techniques for temporal event data often struggle with high-dimensional datasets, limiting effective aggregation.
• Current methods group event types as a pre-process, which imposes significant constraints on analysis flexibility.
• High-dimensionality in event sequence data hinders the ability to identify meaningful patterns and insights.

Purpose of the Study:

• To present a novel visual analytics approach for dynamic hierarchical dimension aggregation.
• To enable interactive exploration of event groupings at runtime, adapting to specific analysis contexts.
• To address the limitations of pre-defined event groupings in high-dimensional temporal data analysis.

Main Methods:

• Developed an algorithm to computationally quantify the informativeness of different grouping levels within a predefined hierarchy at runtime.
• Introduced a scented scatter-plus-focus visualization design with an optimization-based layout for interactive hierarchical exploration.
• Applied the techniques to high-dimensional medical event sequence data.

Main Results:

• The approach allows users to dynamically select the most appropriate level of event grouping during analysis.
• Successfully applied the visual analytics technique to real-world, high-dimensional temporal event data from the medical domain.
• Demonstrated the ability to overcome aggregation challenges posed by high-dimensional event sequence datasets.

Conclusions:

• The proposed dynamic hierarchical dimension aggregation enhances visual analytics for high-dimensional temporal event data.
• Interactive exploration of event groupings provides greater flexibility and insight compared to pre-processing methods.
• The techniques offer a powerful solution for analyzing complex event sequence data across various domains.