Context. Non-conformance in process execution poses a considerable threat to organizations. Conformance issues occur when the actual behavior of a process deviates from the desired one. Such differences can occur because information systems typically support process execution, but do not enforce a particular process. Rather, human interaction drives a process, giving people a certain flexibility in the execution of a particular case. To detect conformance issues, techniques for conformance checking have been introduced [Ca18]. They strive for the automatic detection of deviations between the observed and desired process behavior by comparing recorded event data to a normative process model. Problem. A key assumption of most conformance checking techniques is that all events of a case have timestamps that allow to infer a total order. However, factors like synchronization issues, manual recording of events, and unreliable data sensing, mean that this assumption is often violated in practice. E.g., in healthcare processes, only the date of certain treatments may be known, but not their specific point in time [Ma13]. As a result of this, events are only partially ordered, meaning that a single trace could actually reflect various possible total orders, i.e., different possible resolutions, as depicted in Figure 1. Since different resolutions can have different conformance results, the conformance of such traces becomes uncertain. Contribution. Our work aims to still provide conformance results in the presence of such order uncertainty. Its idea is to use information from the entire event log to estimate the probability of each possible resolution of an uncertain trace. For this estimation, we argue that the probability of a specific resolution may be based on order information derived from similar traces contained in the event log. Intuitively, if a possible resolution denotes an execution order that is frequently observed for other traces, this resolution is expected to be more likely than one that denotes a less common sequence. By incorporating these probabilities in our approach, conformance checking is grounded in a stochastic model.
Auch wenn der Modellbegriff in seinen Ursprüngen erst viel später auftrat, kamen Modelle und Modellierung schon bei den alten Ägyptern zum Einsatz. Bis heute besteht kein eindeutiges Verständnis für das Modellsein altägyptischer Objekte und dessen Definition. Um aber die Objekte in ihrer Funktion als Modell beurteilen zu können sowie den konkreten Nutzen und Zweck ihrer Existenz in Bezug zu ihrem Origin analysieren zu können, bedarf es einer unumstrittenen Grundlage. So können altägyptische Objekte klassifiziert, in ihren Modellfunktionen vergleichbar gemacht und die Sprache der Objekte in ihrem Modellsein entschlüsselt werden.
Well-established design patterns offer the possibility of standardized construction of software systems and can be used in various ways. The systematic use of design patterns in the field of Artificial Intelligence (AI) Systems however, has received little attention so far, despite AI being a popular research area in recent years. AI systems can be used for a wide variety of applications and play an increasingly important role in business and everyday life. AI systems are becoming more complex however, the actual machine learning (ML) task comprises only a small part of the total source code of a system. In order to maintain a clear and structured architecture for such systems and to allow easy maintenance, standardized elements should be reused in the design. This paper describes possible applications of well-known design patterns in AI systems to improve traceability of the system design.
While supporting the execution of business processes, information systems record event logs. Conformance checking relies on these logs to analyze whether the recorded behavior of a process conforms to the behavior of a normative specification. A key assumption of existing conformance checking techniques, however, is that all events are associated with timestamps that allow to infer a total order of events per process instance. Unfortunately, this assumption is often violated in practice. Due to synchronization issues, manual event recordings, or data corruption, events are only partially ordered. In this paper, we put forward the problem of partial order resolution of event logs to close this gap. It refers to the construction of a probability distribution over all possible total orders of events of an instance. To cope with the order uncertainty in real-world data, we present several estimators for this task, incorporating different notions of behavioral abstraction. Moreover, to reduce the runtime of conformance checking based on partial order resolution, we introduce an approximation method that comes with a bounded error in terms of accuracy. Our experiments with real-world and synthetic data reveal that our approach improves accuracy over the state-of-the-art considerably.
Process modeling plays a central role in the development of today’s process-aware information systems both on the management level (e.g., providing input for requirements elicitation and fostering communication) and on the enactment level (providing a blue-print for process execution and enabling simulation). The literature comprises a variety of process modeling approaches proposing different modeling languages (i.e., imperative and declarative languages) and different types of process artifact support (i.e., process models, textual process descriptions, and guided simulations). However, the use of an individual modeling language or a single type of process artifact is usually not enough to provide a clear and concise understanding of the process. To overcome this limitation, a set of so-called “hybrid” approaches combining languages and artifacts have been proposed, but no common grounds have been set to define and categorize them. This work aims at providing a fundamental understanding of these hybrid approaches by defining a unified terminology, providing a conceptual framework and proposing an overarching overview to identify and analyze them. Since no common terminology has been used in the literature, we combined existing concepts and ontologies to define a “Hybrid Business Process Representation” (HBPR). Afterwards, we conducted a Systematic Literature Review (SLR) to identify and investigate the characteristics of HBPRs combining imperative and declarative languages or artifacts. The SLR resulted in 30 articles which were analyzed. The results indicate the presence of two distinct research lines and show common motivations driving the emergence of HBPRs, a limited maturity of existing approaches, and diverse application domains. Moreover, the results are synthesized into a taxonomy classifying different types of representations. Finally, the outcome of the study is used to provide a research agenda delineating the directions for future work.
