Business Process Management and Enterprise Architecture: Friends or Foes?
This is a short write-up of the BPM-EA (Business Process Management and Enterprise Architecture) session of ACM SAC 2018. This track aims at gathering researchers and practitioners around the broad topics of business processes and enterprise architecture with a special interest in modeling. These disciplines are quickly evolving and intertwining with each other, and are often referred to with the broad term of business modeling.
In the following sections, I’m reporting a summary of the presented contributions at the conference.
Ordering Stakeholder Viewpoint Concerns for Holistic Enterprise Architecture (The W6H Framework)
by Mujahid Sultan and Andriy Miranskyy, Ryerson University, Canada
The question here is: can we capture and understand the needs and thoughts of all the stakeholders of the enterprise?
The full presentation is available from the authors here.
When Zachman defined the questions of EA (What, Who, Where, When, How, Why), they called it W5H, starting from journalism. But the questions are:
- Are they enough in the modern enterprise? Do we need more?
- What is the ordering of importance of these questions?
- Can they support iterative and agile methods?
Regarding the first point, the basic linguistic constructs in questioning are: Who, What, Which, and Where. The How follows, and finally Why and When close. So, if we add Which to the W5H, and we get to W6H. The role of Which is that of allowing selectivity, i.e., the possibility of selecting one among a set of options. This was not possible in the standard W5H model.
Regarding the second point: there is actually a logical ordering of the questions, from linguistics and common sense, as shown here:
And not so surprisingly, this maps perfectly to relations between the different aspects of system design, reported in the following figure:
However, Which was not covered in Zachman framework! This is the exended framework with the W6H complete list:
As a last comment: all of this is not very well fitting the agile principles, although the introduction of Which may improve a little along this aspect.
The full presentation is available from the authors here.
Similarity Resonance for Improving Process Model Matching Accuracy
Nour Assy, Boudewijn F. van Dongen, Wil M.P. van der Aalst, Eindhoven University of Technology, The Netherlands
Process Mining is crucial to understand best practices in enterprise. In particular, you want to decide which model is best, for designing a best behaviour vs. observed behaviour vs. all possible behaviours.
As a first step, you need to be able to run process model matching based on similarity. Usually this is done with pair-wise similarity between labels, but this is frequently not so accurate.
The new idea that can be exploited for improving performance, is that similar activities are frequently used in similar contexts.The idea is then to consider similarity of neighbours for calculating the similarity. The steps are:
- Calculate k-nearest neighbours
- Construct k-resonance graph
- Compute similarity resonance
Essentially it’s an eigenvalue calculation problem on the graph.
By applying the method to experimental settings, results show improvement on recall of existing techniques.
Verifying BPMN Understandability with Novice Business Managers
Vanessa Mendoza, Denis Silva da Silveira, Maria Luiza Albuquerque, Federal University of Pernambuco; and Brazil João Araújo, Universidade NOVA de Lisboa, Portugal
The main point of the discussion is to raise the importance and role of model comprehension and understandability in the particular case of business process models, with respect to novice BPM analysts.
It’s well known that BPMN is a complex notation and only a limited part of it is widely used and understood. When comparing textual process definitions with BPMN ones, experimental results show that people that have some knowledge on BPMN understand BPMN better than text, while for people that don’t know BPMN the textual description is largely more understandable.
Using Enterprise Architecture Model Analysis and Description Logics for Maturity Assessment
Diogo Proença and José Borbinha, IST/INESC-ID, Portugal
Maturity models measure the advancement of an organization with respect to a given aspect. Among the most known models we have CMMI maturity model, which defines the levels:
- Initial
- Managed
- Defined
- Quantitatively defined
- Optimizing
Appropriate assessment methods exist for maturity models. The objective is to implement methods and techniques to automate maturity assessment.
The proposed idea is to represent maturity models and EA using ontologies.
Conceptual Modeling of Inter-dependencies between Processes and Data
Carlo Combi, Barbara Oliboni, Mathias Weske, Francesca Zerbato, Università degli Studi di Verona, Italy; Hasso Plattner Institute at the University of Potsdam, Germany
The focus is on the inter-relations between processes and data in organizations. The two aspects are complementar and intertwined, and can be represented with respective models.
The idea is to connect process models and data models by bridging them through a mapping that describe which operations on the data is performed by which activity in the business process.
Based on that, you can detect dependencies and inconsistencies between data and processes, or within processes.
You can also go down to looking into the instance level, namely the data tuples/instances and process traces. This allows to detect actual concrete anomalies in concrete instances.
Automated Analysis of Industrial Workflow-based Models
Mario Cortes-Cornax, Ajay Krishna, Gwen Salaün, University of Grenoble Alpes, France; Adrian Mos, Naver Labs, France
Mangrove is a modeling language, with modeling tools available in Eclipse that store models as DSPML. It represents processes, where wrt BPMN there are no gateways nor any other element except for steps and flows.
This is a Mangrove example model:
It’s possible to run verification passing through a Process Intermediate Format (PIF), which is in turn transformed into a LNT formal model through a model-to-text transformation. This is the metamodel of PIF:
In parallel, it can be transformed into a Z3 model, so that satisfiability can also be tested.
Dynamic High-level Requirements in Self-Adaptive Systems
Davide Rossi, University of Bologna, Italy Francesco Poggi, University of Bologna, Italy Paolo Ciancarini, University of Bologna, Italy
Self-adaptive system automatically align to changing context. It’s particularly hard to describe requirements for these systems.
The idea is to define a semantics-based architecture for requirements-aware self-adaptive systems through modeling and formalization of requirements, including soft-goals. Many different models are defined for the different aspects of the system, and then they are formalized in OWL-2.
Using MOF, Archimate and iStar are mapped to a common modeling framework, together with RDF, OWL and CL through the ODM (Ontology Definition Metamodel) (see figure below).
The selection of reconfiguration policy for adaptation is applied considering the soft goals model, using a simple utility function model. Priorities of goals can dynamically change, as shown in the example below.
Wrapping up..
This was a write-up of the presentations given a the BPM-EA (Business Process Management and Enterprise Architecture) session of ACM SAC 2018.
While several researches are addressing BPM and EA, there is still a strong need of exploring new paths of improvement, integration and consolidation between the two worlds.
Especially in light of the recent trends in the business, which put a lot of emphasis on large-scale system engineering, through modeling techniques that are applied at global scale, and data analysis approaches that combine cloud-based big data analytics with enterprise and system models.
