Process Mining with Context-Aware Slicing:An Initial Approach

Accurate prediction and performance analysis of process executions are key for data-driven decision-making in complex environments. We propose a process mining approach based on context-aware slicing, which segments event data into homogeneous cohorts using static case attributes. The method builds on CRISP-DM and PM{2}HC with a slicing mechanism for comparative analysis. We evaluate it on real-world hospital case, where contextual cohorts exhibit longer durations and lower conformance to reference pathways. For each cohort, we derive performance indicators that can serve as inputs to analytic models. Our results show that context-aware slicing uncovers cohort-specific patterns that may be muted in global averages and enhances interpretability for domain experts.