Visualization of frequent temporal patterns in single or two populations.

Area of Science:

• Data Science
• Clinical Informatics
• Information Visualization

Background:

• Temporal knowledge discovery is vital in clinical data science.
• Progress in computational discovery of frequent temporal patterns exists, but effective visualization remains a challenge.
• Visualizing temporal patterns can significantly aid domain experts in knowledge acquisition.

Purpose of the Study:

• To introduce a novel approach for visualizing frequent temporal patterns, specifically time interval related patterns (TIRPs).
• To facilitate the exploration and comparison of temporal patterns mined from single or multiple populations.
• To assist domain experts in discovering discriminating patterns between populations.

Main Methods:

• Developed a visualization of an enumeration tree for frequent temporal patterns.
• Implemented a bubble chart visualization where axes represent pattern metrics (frequency, reoccurrence).
• Demonstrated the approach on time interval related patterns (TIRPs) and validated with a user study on real-life datasets.

Main Results:

• The novel visualization enables browsing and searching of frequent temporal patterns within an enumeration tree.
• The bubble chart provides a quick overview of patterns and access to specific ones.
• User studies confirmed the usability advantages of the proposed visualization methods for temporal knowledge discovery.

Conclusions:

• The introduced visualization approach enhances the exploration and acquisition of temporal knowledge from clinical data.
• The method is effective for analyzing complex temporal patterns like TIRPs.
• This work offers a significant contribution to the under-researched area of temporal pattern visualization in data science.