Sirvi Autor "Rosenbaum, Jannis" järgi
Nüüd näidatakse 1 - 1 1
- Tulemused lehekülje kohta
- Sorteerimisvalikud
listelement.badge.dso-type Kirje , Optimization of Activity Batching Policies in Business Processes(Tartu Ülikool, 2025) Rosenbaum, Jannis; López-Pintado, Orlenys, juhendaja; Dumas, Marlon, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituutMany business processes group similar tasks into batches to reduce costs and improve efficiency. Batching allows managers to trade off cost and processing effort against waiting time. Larger and less frequent batches can reduce effort by amortizing fixed costs, but they increase waiting times. Conversely, smaller and more frequent batches minimize waiting times but lead to higher processing effort and increased fixed costs. A batching policy defines how tasks are grouped and when each group is executed. This thesis addresses the problem of discovering batching policies that achieve the best balance between waiting time, processing effort, and cost. It proposes a Pareto optimization approach that begins with a given set (possibly empty) of batching policies and generates alternatives using intervention heuristics. Each heuristic identifies an opportunity to improve a policy with respect to a specific metric (e.g., waiting time or resource utilization) and suggests an adjustment. The effects of these interventions are evaluated via simulation. The intervention heuristics are embedded into optimization meta-heuristics — hill climbing, simulated annealing, and reinforcement learning — that iteratively refine the Pareto front of discovered solutions. An experimental evaluation compares the heuristic-driven approach to non-guided baselines, assessing convergence, diversity, and cycle time improvements. The result is a set of optimized batching strategies that present different trade-offs between speed, cost, and resource usage — empowering decision-makers to select the policy that best fits their operational needs. Additionally, a graphical user interface was developed to facilitate intuitive access to the optimizer software and improve the readability of the results.