Business Process Simulation (BPS) plays a central role in analyzing and improving organizational processes by estimating the effects of potential changes. A key element of any BPS model is the case-arrival component, which determines when new cases enter the process. While accurate arrival time modeling is essential for producing reliable simulations, most existing approaches rely on static inter-arrival time distributions that overlook the temporal dynamics inherent in organizational environments, resulting in reduced accuracy and misleading insights. To address this, we propose Auto Time Kernel Density Estimation (AT- KDE), a scalable arrival time modeling approach that captures global trends, weekday effects, and intraday shifts. Across 20 diverse processes, our experiments show that AT-KDE produces more accurate and robust arrival times than existing case-arrival modeling approaches, while maintaining practical execution times. Moreover, we assess how different arrival modeling approaches affect overall simulation quality. Noting that existing BPS evaluation metrics may not explicitly account for process dynamics, we additionally showcase a novel utility-based evaluation framework, which we then use in our experiments.

Top