VAC$^{2}$2: Visual Analysis of Combined Causality in Event Sequences.

Area of Science:

• Data Science
• Complex Systems Analysis
• Causality Research

Background:

• Causality identification is crucial for decision-making in complex systems.
• Existing methods for temporal event sequence data primarily focus on individual causal discovery, neglecting combined causality.
• There is a need for methods that can uncover and represent combined causal relationships within event sequences.

Purpose of the Study:

• To develop a novel approach for discovering and visualizing combined causality in temporal event sequence data.
• To address the limitations of existing methods in capturing the interplay of multiple causes.
• To create an effective visual analysis system for exploring both individual and combined causal factors.

Main Methods:

• Defined eliminating and recruiting principles to balance causality effectiveness and controllability.
• Utilized the Granger causality algorithm with Reactive point processes to model entity interactions.
• Developed a visual analysis system (VAC) employing an "electrocircuit" metaphor for aggregated causality visualization.
• Integrated directed, weighted, and parallel-based hypergraphs with aggregation layout, sorting strategies, and interactive features.

Main Results:

• The VAC system effectively visualizes combined causality with no node-overlap, edge-intersection, or link-ambiguity.
• The system supports multi-level causality exploration using diverse ordering strategies and focus+context techniques.
• Case studies and a user study demonstrated the usefulness and effectiveness of the VAC system in analyzing event sequence data.

Conclusions:

• The developed approach and VAC system successfully address the gap in combined causality discovery for temporal event sequences.
• The "electrocircuit" visualization provides an intuitive and informative representation of complex causal interactions.
• The system empowers users to explore and understand combined causes more effectively, aiding in decision-making and policy implementation.