Discovering Utility-driven Interval Rules

Area of Science:

• Artificial Intelligence
• Data Mining
• Knowledge Discovery

Background:

• High-utility sequential rule mining (HUSRM) identifies event associations in sequences.
• Existing HUSRM methods are limited to point-based sequences, not interval events.
• Interval events are common, and traditional methods struggle to reveal their correlations.

Purpose of the Study:

• To propose a novel algorithm, UIRMiner, for discovering utility-driven interval rules (UIRs).
• To address the limitations of existing HUSRM algorithms in handling interval-event sequences.
• To enable effective knowledge discovery from interval-event sequence databases.

Main Methods:

• Developed the UIRMiner algorithm specifically for interval-event sequences.
• Introduced a numeric encoding relation representation to optimize computation and storage.
• Implemented a complement pruning strategy combining utility upper bound and relation to reduce search space.

Main Results:

• UIRMiner successfully extracts all utility-driven interval rules from interval-event sequence databases.
• The numeric encoding representation significantly reduces computation and storage costs.
• The complement pruning strategy effectively shrinks the search space.

Conclusions:

• UIRMiner is an effective and efficient algorithm for high-utility interval rule mining.
• The proposed methods overcome the complexities of interval-event sequences compared to point-based methods.
• This work advances knowledge discovery in artificial intelligence by enabling analysis of interval-event data.