A novel multi objective grey wolf optimization fuzzy miner for process discovery: Incorporating robustness and...

Area of Science:

• Process mining
• Artificial intelligence
• Metaheuristic optimization

Background:

• Process mining analyzes event logs to understand and improve business processes.
• Existing methods often struggle with noise and lack interpretability.
• There is a need for robust and explainable process discovery techniques.

Purpose of the Study:

• Introduce Fuzzy Multi-Objective Grey Wolf Optimization (Fuzzy MOGWO) for process discovery.
• Simultaneously optimize six key metrics: Fitness, Precision, Generalization, Simplicity, Robustness, and Explainability.
• Evaluate Fuzzy MOGWO's performance against established process mining algorithms.

Main Methods:

• Integration of fuzzy modeling with a multi-criteria metaheuristic optimization approach.
• Development of a normalized scoring mechanism using the L₂ norm for balanced objective evaluation.
• Benchmarking Fuzzy MOGWO against Alpha Miner, Inductive Miner, and Fuzzy Miner on synthetic and real-world event logs, including noisy datasets.

Main Results:

• Fuzzy MOGWO achieved a normalized score of 0.329 in noise-free conditions, outperforming the best baseline by 14.24%.
• In noisy environments, Fuzzy MOGWO scored 0.440, exceeding the top competitor by 16.40%.
• On real-world logs, Fuzzy MOGWO outperformed competitors in 4 out of 6 metrics, demonstrating superior effectiveness and robustness.

Conclusions:

• Fuzzy MOGWO offers a comprehensive and reliable solution for challenging process discovery tasks.
• The proposed method exhibits substantially improved effectiveness, robust performance under noise, and enhanced interpretability.
• Fuzzy MOGWO sets a new standard for multi-objective process discovery by balancing multiple critical performance dimensions.