The research addresses the challenge of scheduling business processes where the exact sequence of activities is uncertain. This uncertainty arises from decisions made during execution based on emerging data. Utilizing probabilistic information derived from historical logs can improve the likelihood of successful completion. The problem is formulated as a chance-constrained optimization problem with two formulations: a decomposed approach and an integrated approach.
The decomposed approach consists of two stages. The planning stage minimizes the expected number of superfluous activities while satisfying a feasibility constraint. The scheduling stage then minimizes makespan over the planned activities. The integrated approach combines planning and scheduling into a single formulation. Evaluation was conducted on two real-world datasets and one synthetic dataset.
Source: https://arxiv.org/abs/2609.05578