Metrics for Process Models

Abstract

PrefaceBusiness process modeling plays an important role in the management of business processes. As valuable design artifacts, business process models are subject to quality considerations. The absence of formal errors such as deadlocks is of paramount importance for the subsequent implementation of the process. This book develops a framework for the detection of formal errors in business process models and for the prediction of error probability based on quality attributes of these models (metrics). We focus on Event-driven Process Chains (EPCs), a widely used business process modeling language due to its extensive tool support. The advantage of this focus is firstly that the results of this book can be directly translated into process modeling practice. Secondly, there is a large empirical basis of models. By utilizing this large stock of EPC model collections, we aim to bring forth general insights into the connection between process model metrics and error probability. In order to validate such a connection, we first need to establish an understanding of which model attributes are likely connected with error probability. Furthermore, we must formally define an appropriate notion of correctness that answers the question of whether or not a model has a formal error. As a prerequisite to answering this question, we must define the operational semantics of the process modeling language formally. ContributionsThis book presents a precise description of EPCs, their control-flow semantics and a suitable correctness criterion called EPC soundness. Furthermore, we identify theoretical arguments on why structural metrics should be connected with error probability and provide an empirical validation of this connection. To be more concise, this book provides the following technical contributions.Formalization of the OR-join: The semantics of the OR-join have been debated for more than a decade. Existing formalizations suffer from either a restriction of the EPC syntax (see [78,247,238,4,101]) or from non-intuitive behavior (see [325,218,11,465]). In Chap. 2, we formalize the EPC semantics concept as proposed elsewhere [267]. In comparison to other approaches this novel formalization has the advantage of not being restricted to a subset of EPCs. Moreover, VIII Preface it provides intuitive semantics for blocks of matching OR-splits and joins since they cannot deadlock. As a proof of concept, we implemented a plug-in for ProM that calculates the reachability graph. In this way, this novel semantics definition contributes to research on the specification of business process modeling languages. Verification of process models with OR-joins and multiple start and end events:Verification techniques for process models with OR-joins and multiple start and end events suffer from one of two problems: Firstly, they build on an approximation of the actual behavior, e.g., by considering a relaxed notion of soundness [101], by involving user decisions [109] or by approximating relaxed soundness with invariants [440]. Therefore, they d...

Improving Business Process Models with Agent-Based Simulation and Process Mining

“…In particular, it is possible to measure the quality of such model through a set of metrics. The use of metrics to evaluate the quality of process models has been thoroughly investigated in the literature [11][12][13][14], but here we focus on a set of metrics that are more tailored to the hierarchical model that is at the core of our approach. Specifically, such hierarchical model should be "balanced" in the sense that all of its components -i.e.…”

“…In the literature, there are several metrics that have been proposed for the analysis of business process models [11][12][13][14]. In some cases, these metrics have their origin in network analysis and software engineering.…”

“…For this purpose, we studied a number of metrics focusing on factors such as size, density, modularity, control-flow, etc. [11,30,31]. The literature provides a lot of metrics to measure these factors, but here we selected a subset of metrics that are more geared towards the structure and complexity of those models.…”

“…of arcs per node [11] NAN Density Ratio of number of arcs to number of nodes. Relational density [32] RD Density Ratio of number of arcs to the total number of possible arcs.…”

Determining the Suitability of a Business Process Modelling Technique for a Particular Application

“…It is suggested that the graphical portrayal of these types of features provides additional support for understanding, implementing, executing, measuring, analysing, improving, and controlling business processes. According to Mendling, business process models can be considered to be the conceptual artefacts underpinning the management of organisational processes [1]. In turn, BPM has evolved into a valuable approach to defining and documenting business processes [2], which are considered an important prerequisite for organisations that intend to engage in business process improvement and management initiatives [3].…”