Exploring Multivariate Event Sequences Using Rules, Aggregations, and Selections

Area of Science:

• Computer Science
• Data Analysis
• Information Visualization

Background:

• Multivariate event sequences are common in various domains like travel, telecommunications, and server logs.
• Existing analysis methods often focus on either temporal or structural aspects of event data, but not both simultaneously.
• Events possess not only temporal information but also associated multivariate data, requiring integrated analysis.

Purpose of the Study:

• To develop an integrated approach for exploring multivariate event sequences at both multivariate and sequential levels.
• To enable users to define and apply rewrite rules using multivariate regular expressions for pattern discovery.
• To facilitate interactive enrichment and simplification of event sequences for deeper investigation.

Main Methods:

• A novel approach allowing simultaneous exploration of multivariate and sequential event data.
• Interactive definition of rewrite rules using multivariate regular expressions.
• A bottom-up, glyph-oriented approach (Eventpad) for searching, clustering, and aligning event sequences.
• Storing identified patterns as new event types or attributes for further analysis.

Main Results:

• Demonstrated effectiveness of the approach using real-world datasets, including telecommunication traffic and hospital treatments.
• Enabled users to simultaneously analyze temporal and multivariate attributes of event sequences.
• Facilitated the discovery and utilization of domain-specific patterns within complex event data.

Conclusions:

• The proposed approach effectively addresses the limitations of existing methods for analyzing multivariate event sequences.
• Eventpad provides a powerful tool for interactive exploration and pattern discovery in complex, real-world event data.
• Simultaneous analysis of multivariate and sequential data enriches the understanding and simplification of event sequences.