Elsevier

Computers in Industry

Volume 67, February 2015, Pages 72-85
Computers in Industry

Mining version histories for change impact analysis in business process model repositories

https://doi.org/10.1016/j.compind.2014.10.005Get rights and content

Highlights

  • We propose a novel approach to change impact analysis in an organization's repository of business process models.

  • Our approach mines the revision history of a process model repository and uses this knowledge for change impact estimation.

  • The result of an evaluation using the version history of a real business process repository indicated the effectiveness of our approach and its outperforming a basic dependency-based impact analysis.

Abstract

In order to remain competitive and sustainable in today's ever-changing business environments, organizations need to frequently make changes to their business activities and the corresponding business process models. One of the critical issues that an organization faces is change impact analysis: estimating the potential effects of changing a business process to other processes in the organization's business process repository. In this paper, we propose an approach to change impact analysis which mines a version history of a business process model repository. Our approach then identifies business process models that have been co-changed in the past and uses this knowledge to predict the impact of future changes. An empirical validation on a real business process model repository has showed the effectiveness of our approach in predicting impact of a change.

Introduction

A business process is defined as consisting of a set of activities, performed by their relevant roles or collaborators, to intentionally achieve a set of common business goals [1]. Business processes are the core assets of any enterprise, covering many aspects in industry such as design, engineering, manufacturing, purchasing, physical distribution, production management and supply chain management. Organizations committed to long-term business process management (BPM) may have repositories of hundreds or even thousands of business process models. For example, the IBM BIT Process Library has 735 process models [2], the SAP Reference Model contains 604 process models [3], and there are 6000+ process models in Suncorp's process model repository for insurance [4]. On the other hand, the ever-changing business environment (due to various reasons such as new customer requirements, global competition pressures, new regulations, new IT solutions, economic down turn, etc.) demands organizations to constantly consider changing their business activities in order to remain competitive and sustainable in the long term.

Business process models are essential knowledge assets for an organization (with hundreds and thousands of business process models [5]) to manage its business processes in terms of documenting and implementing procedures, control their execution, analyze their performance, and improve them (i.e. business process management [6], [7]). Recent studies (e.g. [8]) have demonstrated various perceived benefits of using business process models as the basis for process improvement, understanding, communication, execution, analysis and simulation. In particular, both the practitioner and vendor groups participating the study conducted in [8] agreed that support for continuous improvement of an organization's business processes (in order to react to changes in its business environment) is the core benefit of business process models. Changes in process models need to be put into practice immediately in order for process improvement to be effective [1]. Recent work (e.g. [9]) provide techniques for automatic execution of business process models or develop process engines that can interpret process models and enact them automatically. The explicit documentation of business processes in those contexts facilitates change management in terms of quickly identifying what needs to be changed and implementing those changes rapidly.

However, making changes to large, complex repositories of business process models is a highly challenging task. This is mainly due to the ripple effect caused by a change. Specifically, a change made to one business process can potentially affect a range of other processes that are related to the process being changed. For example, changes initially made to a sub-process (e.g., adding a new activity) may lead to secondary, additional changes made to the processes that contain this sub-process. Such changes made to those processes may lead to further changes in other related processes. In a large repository of hundreds or even thousands of business process models, it becomes critical to determine the impact of a change, which is the core focus of our paper.

Change impact analysis usually starts with the business analyst examining the change request and determining the processes initially affected by the change (i.e. the primary changes). The business analyst then determines other process models in the business process repository that are potentially affected and required to be changed. Changes made to those impacted processes may also potentially affect other processes and thus the impact analysis continues this procedure until a complete impact set is obtained. Change impact analysis plays a major part in planning and establishing the feasibility of a change in terms of predicting the cost and complexity of the change (before implementing it). This helps reduce the risks associated with making changes that have unintended, expensive, or even disastrous effects on existing business operations.

Organizations face a range of challenges in managing change in the context of large and complex collections of business processes. While an important class of change management problems pertain to process instances, our focus in this paper is however exclusively on process designs/models. Here, we take a model-based approach to impact analysis which examines impacts to the business processes before the implementation of such changes. An appropriate decision can therefore be made (before any detailed implementation of the change is considered) on whether to implement a specific set of changes based on what business process models are likely to be impacted and thus on the likely change cost. Earlier decision making and change planning are clearly important in the context of rigorous change management. We also acknowledge that there may be a gap between the actual execution of a process and what is being described in its model, and change impact analysis for process instances is an alternative.

Although business process management research is gaining increasing attention from both industry and academia, there has been very little work on supporting change impact analysis in business process model repositories [5]. Some recent work (e.g. [10]) only focus on identifying dependency relationships among different entities in a single process in order to analyze the impact of a change made to one part of the process to other parts of the process. The work in [11] specifically aims to support change impact analysis between services and business processes in a service-oriented environment. The recent work in [12] addresses the issue of propagating changes to maintain consistency within a process repository, which is part of change implementation rather than change impact analysis.

Traditional approaches to change impact analysis in business processes tend to focus on establishing inter-process relationships (i.e. dependencies) and using this knowledge for impact analysis. These approaches rely on a classification of relationships between business processes (e.g., [13], [12]). However, basic dependency-based impact analysis techniques are considered to be conservative in that they consider all possible transitive closure of inter-process relationships. Results produced by those techniques may have enormous impact sets, which are sometimes unnecessary or even too large to be of practical use [14]. In addition, establishing inter-process relationships in such a way that precisely reflect the semantic nature of dependency between processes (e.g., annotating process models with semantic effects as done in [15]) may be labour-intensive and time consuming.

In this paper, we propose an alternative approach which focuses on detecting factual inter-process dependencies as manifested in the evolution of the process models. Our approach mines the revision history of a process model repository, and identifies processes that have been frequently changed at the same time to identify co-variation patterns between them. This approach computes the impact based on the heuristic that processes that have been changed together in the past (co-variation coupled) will be likely changed together in future. We have performed an empirical validation using a real business process repository to compare the effectiveness of our approach against the basic dependency-based impact analysis technique.

The structure of this paper is as follows. In the next section, we briefly describe how business processes are defined using the standard Business Process Modeling Notation and how they are annotated with semantic effects. In Section 3, we present a generic framework for change impact analysis in business process repositories. We then discuss in detail a basic inter-process relationship (dependency-based) analysis approach (Section 4) and our revision history mining approach (Section 5). We report on our evaluation in Section 6 and discuss related work in Section 7. We then conclude and outline our future work in Section 8.

Section snippets

Business Process Modeling Notation

While there are a range of modeling notations for business processes, for our purposes we use the Business Process Modeling Notation (BPMN) which is a standard for business process modelling [16]. It provides graphical notation for specifying various types of activities, decision responsibilities, control and data flow in business process within one organization and in cross-organizational settings. BPMN has been widely used in the industry due to its powerful notation which is readily

Change impact analysis framework

In this section, we describe a framework for change impact analysis in a process repository (see Fig. 3). Impact analysis starts with the business analyst examining the change request, identifying the processes in the current process repository (i.e., PRcurrent in Fig. 3) initially affected by the change, and possibly making changes to those processes. Such primary changes yield a new version PRmodified of the process repository. Our framework then compares the two versions and automatically

Dependency-based impact analysis

This approach focuses on identifying relationships between processes in a process repository. The semantic effects can be used to establish semantic relationships between processes. For this baseline approach, we used four types of inter-process relationships: part-whole, inter-operation, generalization-specialization [12], and data-dependent. We briefly describe them as below.

  • Part-whole: this type of relationship exists between two processes when one process is required by the other to fulfill

Mining revision history approach

The above approach requires all business process models in a process repository to be fully annotated with semantic effects. To do so, each activity in each process model needs to be annotated with immediate effects. This is highly time-consuming and expensive, especially with large process repositories of hundreds or even thousands of processes. For example, the IBM BIT Process Library [2] contains 735 process models, each of which has on average 27 activities, which requires 19,845 annotation

Evaluation

An important question that we would like to address in this evaluation is: how well the mining revision history approach works, compared with the traditional, basic inter-process relationship approach. We would like to understand whether the mining revision history approach can offer similar accurate and precise impact analysis to the inter-process relationship approach, but without the expensive cost of annotating business process models. We now discuss how we have designed such an experiment,

Related work

In the past few years, dealing with large repositories of business models has attracted an increasing attention from both industry and academia since it has become more common see organizations dealing with repositories of hundreds or thousands business process models [5]. Various techniques which address different aspects of managing process model repositories have been proposed in literature such as querying for a particular process model (e.g. [35]), searching for similar process models

Conclusions and future work

In this paper, we have described two distinct approaches to change impact analysis in business process models. The basic dependency-based approach requires semantic annotation of business process model but does not require revision histories. On the other hand, the mining approach (which is our main contribution) does not require semantic annotation. Our mining approach uses the revision history of a process repository to predict change impact based on the assumption that processes that were

Hoa Khanh Dam is Lecturer in Software Engineering at the School of Computer Science and Software Engineering, University of Wollongong, Australia. He holds PhD and Master degrees in Computer Science from RMIT University and a Bachelor of Computer Science degree from the University of Melbourne in Australia. His research interests include software engineering, multi-agent systems, service-oriented computing and business process management. He served as Program Co-Chair for the 17th

References (57)

  • M. La Rosa et al.

    APROMORE: an advanced process model repository

    Expert Systems with Applications

    (2011)
  • H. Smith et al.

    Business Process Management: The Third Wave

    (2006)
  • D. Fahland et al.

    Instantaneous soundness checking of industrial business process models

  • T. Curran et al.

    SAP R/3 Business Blueprint: Understanding the Business Process Reference Model

    (1998)
  • M. La Rosa et al.

    Business process model merging: an approach to business process consolidation

    ACM Transactions on Software Engineering and Methodology

    (2013)
  • W.M.P. van der Aalst

    Business process management: a comprehensive survey

    ISRN Software Engineering

    (2013)
  • M. Weske

    Business Process Management: Concepts, Languages, Architectures

    (2012)
  • M. Indulska et al.

    Business process modeling: perceived benefits

  • Y. Wang et al.

    A change analysis tool for service-based business processes

  • T.A. Kurniawan et al.

    Relationship-preserving change propagation in process ecosystems

  • T.A. Kurniawan et al.

    On formalizing inter-process relationships

  • S. Lehnert

    A taxonomy for software change impact analysis

  • K. Hinge et al.

    Process seer: a tool for semantic effect annotation of business process models

  • Object Management Group, Business Process Model and Notation (BPMN), v1.2, http://www.omg.org/spec/BPMN/1.2...
  • D. Fensel et al.

    Web service modeling ontology

    Semantic Web Services

    (2011)
  • M. Hepp et al.

    Semantic business process management: a vision towards using semantic Web services for business process management

  • I. Di Pietro et al.

    Model checking semantically annotated services

    IEEE Transactions on Software Engineering

    (2012)
  • F. Smith et al.

    Rule-based behavioral reasoning on semantic business processes

  • Cited by (14)

    • Recalling the rationale of change from process model revision comparison – A change-pattern based approach

      2017, Computers in Industry
      Citation Excerpt :

      In the next steps change patterns are checked whether corresponding pre- and postconditions exist within the original revision and the final revision that prove the actual application of a change. An approach that makes use of revision histories of process models to analyze relations between changes and assess their impact is presented by [22]. Similar to our approach their mining approach does not require semantic annotation although they use detected changes for a different purpose.

    • Impact analysis and change propagation in service-oriented enterprises: A systematic review

      2015, Information Systems
      Citation Excerpt :

      However execution trace analysis in service-based systems is extremely challenging due to the distributed and heterogeneous nature of such systems involving different software, hardware architectures, protocol implementations and unstructured service execution data [3]. History mining leverages Mining Software Repositories (MSR) techniques to uncover historical co-change dependencies [12,40,94]. This approach primarily focuses on version histories of process models and Web services.

    • Change propagation method for process variants to configurable process models

      2022, Jisuanji Jicheng Zhizao Xitong/Computer Integrated Manufacturing Systems, CIMS
    View all citing articles on Scopus

    Hoa Khanh Dam is Lecturer in Software Engineering at the School of Computer Science and Software Engineering, University of Wollongong, Australia. He holds PhD and Master degrees in Computer Science from RMIT University and a Bachelor of Computer Science degree from the University of Melbourne in Australia. His research interests include software engineering, multi-agent systems, service-oriented computing and business process management. He served as Program Co-Chair for the 17th International Conference on Principles and Practice of Multi-Agent Systems.

    Aditya Ghose is Professor of Computer Science at the University of Wollongong. He leads a team conducting research into knowledge representation, agent systems, services, business process management, software engineering and optimization and draws inspiration from the cross-fertilization of ideas from this spread of research areas. He works closely with some of the leading global IT firms. Ghose is President of the Service Science Society of Australia and served as Vice-President of CORE (2010–2014), Australia's apex body for computing academics. He holds PhD and MSc degrees in Computing Science from the University of Alberta, Canada (he also spent parts of his PhD candidature at the Beckman Institute, University of Illinois at Urbana Champaign and the University of Tokyo) and a Bachelor of Engineering degree in Computer Science and Engineering from Jadavpur University, Kolkata, India.

    View full text