Prescriptive Process Monitoring Under Resource Constraints: A Causal Inference Approach

Abstract Prescriptive process monitoring is a family of techniques to optimize the performance business by triggering interventions at runtime. Existing prescriptive assume that number may be triggered unbounded. In practice, though, consume resources with finite capacity. For example, in loan origination process, an intervention consist preparing alternative offer increase applicant’s chances taking loan. This requires time from credit officer. Thus, it not possible trigger this all cases. paper proposes technique triggers cost function under fixed resource constraints. The relies on predictive modeling identify cases are likely lead negative outcome, combination causal inference estimate effect case’s outcome. These estimates used allocate maximize function. A preliminary evaluation suggests approach produces higher net gain than purely (non-causal) baseline.