SCOPE: Sequential Causal Optimization of Process Interventions

TL;DR

SCOPE addresses the need to optimize sequences of interventions in PresPM by marrying backward induction with causal learning, allowing policies to be learned directly from observational event logs without process simulators. It models sequential decisions through Q- and V-functions and estimates them with S-, T-, or RA-learners, enabling aligned decisions across all decision points. Empirical evaluation on synthetic and semi-synthetic data shows SCOPE consistently outperforming state-of-the-art sequential PresPM approaches, and provides a reusable semi-synthetic benchmark for future research. The work demonstrates the practical potential to improve KPI outcomes while mitigating bias from process approximations and simulation-based training.

Abstract

Prescriptive Process Monitoring (PresPM) recommends interventions during business processes to optimize key performance indicators (KPIs). In realistic settings, interventions are rarely isolated: organizations need to align sequences of interventions to jointly steer the outcome of a case. Existing PresPM approaches fall short in this respect. Many focus on a single intervention decision, while others treat multiple interventions independently, ignoring how they interact over time. Methods that do address these dependencies depend either on simulation or data augmentation to approximate the process to train a Reinforcement Learning (RL) agent, which can create a reality gap and introduce bias. We introduce SCOPE, a PresPM approach that learns aligned sequential intervention recommendations. SCOPE employs backward induction to estimate the effect of each candidate intervention action, propagating its impact from the final decision point back to the first. By leveraging causal learners, our method can utilize observational data directly, unlike methods that require constructing process approximations for reinforcement learning. Experiments on both an existing synthetic dataset and a new semi-synthetic dataset show that SCOPE consistently outperforms state-of-the-art PresPM techniques in optimizing the KPI. The novel semi-synthetic setup, based on a real-life event log, is provided as a reusable benchmark for future work on sequential PresPM.

Figures (4)

• Figure 1: Gain on SimBank across confounding levels ($\delta$) for different training sizes. The shaded area shows one standard error over 10 iterations. SCOPE and Sep: S-learner with XGBoost.
• Figure 2: Gain on SimBank across confounding levels ($\delta$) for different learners. The shaded area shows one standard error over 10 iterations. SCOPE and Sep: XGBoost, trained on 10K cases.
• Figure 3: Gain on SimBank across confounding levels ($\delta$) for different base models. The shaded area shows one standard error over 10 iterations. SCOPE and Sep: S-learner trained on 10K cases.
• Figure 4: Gain on SimBPIC17 across varying numbers of decision points for three confounding levels ($\delta$). The shaded area shows one standard error over 10 iterations. SCOPE and Sep: S-learner with XGBoost, trained on 10K cases.

Theorems & Definitions (2)

• definition 1: Event, Event Log, Trace, Prefix
• definition 2: Decision point, Intervention action, Policy, Outcome