Preference-Aware Process Analytics for Enterprise Operations: Event-Log Mining, Multi-Criteria Model Selection, and Business Performance Interpretation
Main article
Abstract
Enterprise process analytics has moved from descriptive dashboarding toward decision-oriented interpretation of event logs. However, process discovery still often depends on default algorithm settings, informal preprocessing decisions, and technical quality scores that are difficult for managers to translate into business action. This article develops a preference-aware process analytics framework for enterprise operations. The framework combines event-log mining, candidate discovery pipelines, multi-criteria model selection, and business performance interpretation in a single analytical workflow. Instead of treating fitness, precision, simplicity, generalization, stability, and computational cost as isolated technical indicators, the proposed approach converts them into weighted preference profiles that reflect operational priorities such as compliance, lead-time reduction, service reliability, cost-to-serve control, and managerial interpretability. A numerical study using three calibrated enterprise event-log settings - order-to-cash, procure-to-pay, and service-desk resolution - demonstrates how different preference structures change the recommended process model and alter subsequent business conclusions. The results show that balanced preference-aware selection improves composite interpretation scores by 13.8% relative to default discovery, reduces analyst review time by 25.6% to 33.1% across the three settings, and improves the consistency between discovered process variants and business key performance indicators. Sensitivity analysis further indicates that precision-oriented preferences protect compliance interpretation but may reduce simplicity and managerial readability, while efficiency-oriented preferences improve runtime but may hide low-frequency exceptions. The study contributes to business data analytics by reframing process discovery as a managerial decision problem rather than a purely technical modeling problem. It also offers practical guidance for organizations seeking to govern process mining projects, document selection assumptions, and connect event-log evidence to performance improvement decisions.
