Leveraging Dual Timestamps to Enhance Process Model Construction Using Time-Infused Inductive Miner

In business process management, the increasing complexity of modern business processes underscores the need for accurately understanding parallel activities to efficiently allocate and utilize limited corporate resources. However, many existing event log analysis algorithms focus solely on the sequence of executed activities, limiting their ability to model processes involving simultaneous activities effectively. This research seeks to enhance the inductive miner, which is known for its flexibility, formal guarantees, and scalability, by enabling it to better capture and represent the complexities of real-world business processes. The core modification involves leveraging dual timestamps, which utilize both the start and end times of activities. This approach enhances the algorithm’s ability to understand parallel structures within process models. Through experimentation, limitations of single timestamp were identified, including reduced accuracy in modeling parallel activities. The proposed approach also demonstrates the capability to effectively minimize dummy activities when discovering process models containing both sequential and parallel activities. Experimental results highlight the improved performance of the proposed algorithm, with metrics showing a fitness score of 0.815, precision of 0.827, generalization of 0.846, and simplicity of 0.834. Moreover, the algorithm achieves the highest conformance results with fewer traces required compared to the alpha miner, fuzzy miner, heuristic miner, and the original inductive miner.