Targeted Mining of Time-Interval Related Patterns

Area of Science:

• Data Mining
• Pattern Recognition
• Computational Science

Background:

• Sequential pattern mining often overlooks event durations, treating temporal events as single points.
• Time-interval-related pattern (TIRP) mining addresses this by considering event durations, with applications in healthcare and finance.
• Mining all possible TIRPs is computationally intensive and resource-demanding.

Purpose of the Study:

• To propose a novel algorithm, TaTIRP, for discovering targeted time-interval-related patterns (TIRPs).
• To enhance the efficiency and accuracy of TIRP mining by focusing on specific criteria.
• To improve data analysis for applications requiring consideration of event durations.

Main Methods:

• Development of the TaTIRP algorithm for targeted TIRP discovery.
• Implementation of multiple pruning strategies to eliminate redundant computations.
• Evaluation on diverse real-world and synthetic datasets.

Main Results:

• TaTIRP demonstrates effective discovery of targeted TIRPs.
• Pruning strategies significantly enhance performance on large-scale datasets.
• Experimental results validate the algorithm's accuracy and efficiency.

Conclusions:

• TaTIRP offers an efficient approach to mining time-interval-related patterns.
• Targeted mining improves data analysis efficiency and relevance.
• The algorithm provides a valuable tool for temporal data analysis, especially in fields like healthcare and finance.