Soft Patterns: Moving Beyond Explicit Sequential Patterns During Visual Analysis of Longitudinal Event Datasets

Abstract

Reflecting the growing importance of longitudinal and sequential event analysis to many fields, researchers have proposed a variety of visualization methods and interactive visual analysis techniques in recent years. Early work focused on visualizing all permutations, and approach which scales poorly when applied to large-scale datasets with many event types and long sequence lengths. More recently, a variety of proposed techniques use methods such as simplification algorithms, manual user-driven simplification interaction capabilities, pattern mining algorithms, and statistical prioritization to overcome these scalability challenges. However, these methods commonly seek explicit sequential patterns of the form ‘Event A before Event B’ subject to certain time constraints (e.g., within 60 days) or boolean operators (e.g., ‘Event A or Event B’). This paper asserts that this approach is not sufficient for certain high-dimensional real-world event analysis applications. Instead, it argues that new research is needed to develop methods and tools that support soft pattern discovery and analysis.

Publication
IEEE VIS Workshop on Temporal and Sequential Event Analysis