Clear Visual Separation of Temporal Event Sequences

TL;DR

This paper introduces methods for automatically learning composite events and linked visualizations to improve the interpretability and analysis of complex temporal event sequences, addressing challenges of data volume, variety, and timing.

Contribution

It presents novel techniques for composite event learning and visualization that enhance the analysis of temporal sequences compared to traditional pattern mining methods.

Findings

Composite event learning improves sequence aggregation.

Linked visualizations help identify critical flows.

Metrics guide relevant pattern selection.

Abstract

Extracting and visualizing informative insights from temporal event sequences becomes increasingly difficult when data volume and variety increase. Besides dealing with high event type cardinality and many distinct sequences, it can be difficult to tell whether it is appropriate to combine multiple events into one or utilize additional information about event attributes. Existing approaches often make use of frequent sequential patterns extracted from the dataset, however, these patterns are limited in terms of interpretability and utility. In addition, it is difficult to assess the role of absolute and relative time when using pattern mining techniques. In this paper, we present methods that addresses these challenges by automatically learning composite events which enables better aggregation of multiple event sequences. By leveraging event sequence outcomes, we present appropriate linked visualizations that allow domain experts to identify critical flows, to assess validity and to understand the role of time. Furthermore, we explore information gain and visual complexity metrics to identify the most relevant visual patterns. We compare composite event learning with two approaches for extracting event patterns using real world company event data from an ongoing project with the Danish Business Authority.