Orientation and conformance: A HMM-based approach to online conformance checking. Issue 102 (December 2021)

• Record Type: Journal Article
• Authors: Lee, Wai Lam Jonathan; Burattin, Andrea; Munoz-Gama, Jorge; Sepúlveda, Marcos
• Is Part Of: Information systems. Issue 102(2021)
• Publication Date: 2021-12
• Subjects: Conformance checking -- Online processing -- Business process management; Database management -- Periodicals; Electronic data processing -- Periodicals; Bases de données -- Gestion -- Périodiques; Informatique -- Périodiques; Database management; Electronic data processing; Periodicals; 005.7
• DOI: 10.1016/j.is.2020.101674
• Languages: English
• ISSNs: 0306-4379

Abstract

Abstract: Online conformance checking comes with new challenges, especially in terms of time and space constraints. One fundamental challenge of explaining the conformance of a running case is in balancing between making sense at the process level as the case reaches completion and putting emphasis on the current information at the same time. In this paper, we propose an online conformance checking framework that tackles this problem by incorporating the step of estimating the "location" of the case within the scope of the modeled process before conformance computation. This means that conformance checking is broken down into two steps: orientation and conformance. The two steps are related: knowing "where" the case is with respect to the process allows a conformance explanation that is more accurate and coherent at the process level and such conformance information in turn allows better orientations. Based on Hidden Markov Models (HMM), the approach works by alternating between orienting the running case within the process and conformance computation. An implementation is available as a Python package and experimental results show that the approach yields results that correlate with prefix alignment costs under both conforming and non-conforming scenarios while maintaining constant time and space complexity per event. Highlights: An online conformance checking technique is proposed. Proposed technique balances between making sense at the process level and putting emphasis on the current information. Proposed technique maintains constant time and space complexity per event given the process model. Experimental results on benchmark synthetic dataset shows a F1-score of 0.923 in the classification of whether if a case is conforming. Experimental results on a real-life dataset identifies specific conformance issues.