=Paper=
{{Paper
|id=Vol-2196/BPM_2018_paper1
|storemode=property
|title=Exploring Complexity Metrics for
Artifact-Centric Business Process Models
|pdfUrl=https://ceur-ws.org/Vol-2196/BPM_2018_paper_1.pdf
|volume=Vol-2196
|authors=Mike A. Marin
|dblpUrl=https://dblp.org/rec/conf/bpm/Marin18
}}
==Exploring Complexity Metrics for
Artifact-Centric Business Process Models==
https://ceur-ws.org/Vol-2196/BPM_2018_paper_1.pdf
Exploring Complexity Metrics for
Artifact-Centric Business Process Models
Mike A. Marin1,2
1
University of South Africa
330 Preller St, Muckleneuk, Pretoria, 0002, South Africa
2
IBM Corporation, Hybrid Cloud
3565 Harbor Blvd, Costa Mesa, CA 92626, U.S.A
mmarin@acm.org
Abstract. This study explores complexity metrics for business artifact
process models described by Case Management Model and Notation
(CMMN). Over the last few decades business artifacts with Guard-Stage-
Milestone (GSM) has emerged in the Business Process Management
(BPM) literature as a new way of describing data intensive business
processes. The work on GSM influenced the CMMN standard, which was
created to fill a market need for more flexible case management processes
for knowledge workers.
Complexity metrics have been developed for traditional BPM models,
such as the Business Process Model and Notation (BPMN). However,
traditional BPM is not suitable for describing GSM or CMMN process
models. Therefore, complexity metrics developed for traditional process
models may not be applicable to business artifact process models such as
CMMN. This research addresses this gap and advances literature in the
areas of method complexity, complexity metrics for process models, declar-
ative processes, and research on CMMN by characterizing CMMN method
complexity, identifying complexity metrics for CMMN, and exploring the
relationship between CMMN and GSM.
Keywords: Business Artifacts · Business Process Management · BPM ·
Business Process Model · Case Management · Case Management Model
and Notation · CMMN · Guard-Stage-Milestone · GSM · Complexity
Metric · Process Model Complexity · Method Complexity
1 Introduction
The purpose of this study was to explore complexity metrics for artifact-centric
business process models. Since 1996, after Daneva et al. [2] published the first
report on business process model complexity, research on the topic has focused on
imperative process models. Opportunities for research into data-centric processes
were opened with the introduction of case handling by van der Aalst and Berens
[7]. In addition, the introduction of Business Artifacts (BA) by Nigam and Caswell
[5] expanded data-centric research opportunities into declarative processes.
2 Marin
The publication of Case Management Model and Notation (CMMN) [6] by the
Object Management Group (OMG) in 2014 introduced a standard specification
for declarative processes. However, complexity metrics for declarative process
models have not as of yet been studied. To close this gap, this study explores
complexity metrics for the artifact-based declarative case management modeling
standard, CMMN.
1.1 Motivation
Extensive research has been conducted on the complexity metrics of imperative
process modeling languages. Between 1996 and 2015, starting with the work
produced by Daneva et al. [2] and ending with that done by Kluza [3], more
than a hundred metrics for imperative process modeling languages were proposed.
However, as far as this researcher is aware, no research has been done on the
complexity metrics for declarative process model languages.
With the introduction of the CMMN standard based on GSM the use of
declarative processes is becoming mainstream, and the need for complexity metrics
will increase. Unless research on complexity metrics for declarative processes is
advanced our knowledge of process technology complexity will be incomplete. A
better understanding of complexity metrics for declarative processes will improve
our understanding of these processes, and will help vendors and users of the new
CMMN standard.
1.2 Objectives and Contributions
The main goal of this study were to identify and validate complexity metrics for
artifacts based process models. The study addressed the following objectives:
1. To formalize CMMN as the basis to identify metrics.
2. To formalize the relationship between CMMN and GSM.
3. To assess the method complexity of CMMN.
4. To analyze the applicability of Business Process Management (BPM) com-
plexity metrics to CMMN.
5. To identify complexity metrics for CMMN.
6. To validate the identified complexity metrics.
This research makes contributions in the areas of formalizing CMMN using
first-order logic, proposing bidirectional transformations between CMMN case
types and GSM artifact types, proposing complexity metrics for CMMN, iden-
tifying and comparing the model complexity of CMMN against other process
notations, and clarifying the relationship between CMMN and GSM.
2 Method Complexity and Transformations
The calculated method complexity of CMMN seems to indicate that it is more
complex than Event-driven Process Chain (EPC), but less complex than Business
Exploring Complexity Metrics for Artifact-Centric Process Models 3
Process Management and Notation (BPMN). This may also indicate that BPMN
is more expressive than CMMN. The results are encouraging as they may indicate
that CMMN should be easier to learn than BPMN.
This study also contributes formal transformations between CMMN case types
and GSM artifact types to the broader BA knowledge base. The transformation
from a GSM artifact type to a CMMN case type is relatively straightforward
and simple to describe. The resulting case type modeled using CMMN is visually
similar to the original artifact type, allowing for it to be easily understood
by human beings. The transformation of a case type into an artifact type is
more complex because CMMN extends GSM by introducing new constructs and
defining a life cycle with a set of standard events for those constructs.
3 Systematic Literature Review
A systematic literature review (SLR) was conducted. The review was designed
to identify complexity metrics for process models that have been proposed in the
last 20 years (from January 1996 to June 2016 inclusive), and how these were
validated. The goal of this review was to identify complexity metrics for process
models with the specific purpose of identifying metrics that could be relevant to
CMMN. In addition, it was conducted in order to identify the research methods,
present in the literature about process models, used to validate complexity metrics
that could be adapted to validate CMMN complexity metrics. The principal
findings of this review were follows:
– Large number of proposed metrics without validation.
– No complexity metrics for declarative process models have been proposed.
– There is no agreement on research methods for empirical validation.
– Some consensus on theoretical validation is emerging.
– Insufficient guidance on sample size for empirical validation was uncovered.
– Low rigor and relevance of most empirical validations.
4 Metrics Validation
The validation of complexity metrics for business processes requires both theo-
retical and empirical validation. This study defined a set of metrics for CMMN.
The metrics were theoretically validated using the formal framework for software
measurements as defined by Briand et al. [1] and the complexity metrics were
further validated using Weyuker’s properties [8] for software complexity measures.
This research attempted an approach to empirically validate the metrics based
on pairwise comparisons. This study departs from most empirical validations of
complexity metrics for process models in several important ways, including the
use of power calculations during the experimental design phase, the study targets
professional process modelers instead of student, and it uses a within-subjects
pairwise comparison experimental design.
4 Marin
4.1 Methodology
The objective of the empirical validation was to compare the calculated complexity
metrics against human perceived complexity. This research adopted a quantitative
approach using an online survey [4]. The research question was: does calculated
complexity correlate with human perceived complexity?
This research hypothesized a positive correlation between calculated com-
plexity and perceived complexity, a negative correlation between calculated
complexity and model comprehension, and a negative correlation between model
comprehension and perceived complexity.
4.2 Experimental Design
The research methodology followed a within-subjects experimental design. The
subjects consisted of a convenient sample of professional process modelers sourced
through a set of online forums and mailing lists.
The experiment was implemented using an online survey and tutorial. A
short tutorial3 was included, because the target population might not have been
familiar with CMMN. The experiment used a repeated measures design with
counterbalancing where each subject was exposed to only two models.
4.3 Findings
Although careful planning, sound experimental design, and efforts to minimize
threats to validity were in place, the experiment experienced some problems. In
retrospect, the experiment was too complex and required too many subjects to
be conducted using an unsupervised online survey.
No evidence was found to support any of the four complexity metrics. This
experiment failed to reject the five null hypotheses. Therefore, more research is
required to re-test those hypotheses with smaller and more targeted experiments.
Evidence that a metric that was not included in this experiment, for method-
ological reasons, may predict CMMN model complexity was uncovered. That
particular metric was kept constant at 90. The results obtained (inability to
reject the five null hypotheses) are consistent with all of the models being equally
complex. A post-hoc test designed to identify if subjects found all of the models
equally complex, seemed to indicate that there is evidence to support this new
hypothesis. However, more research will be required to test this new hypothesis.
5 Conclusion
The purpose of this research was to explore complexity metrics for artifact-centric
business process models. The results of the SLR that was conducted for this thesis
confirmed that to date, all of the proposed complexity metrics for process models
have been for imperative process models rather than for artifact-centric business
3
Link to CMMN Tutorial at http://cmmn.byethost4.com
Exploring Complexity Metrics for Artifact-Centric Process Models 5
process models. This research was based on the assumption that complexity
metrics for declarative process models can be derived from the research that has
been conducted on complexity metrics for imperative process models. This thesis
focused on CMMN because of the potential practical impact that such research
may have on vendors and users of this emerging standard.
The goal was to use CMMN as a proxy to fill the gap in the literature
concerning complexity metrics for declarative process models. As such, this
research contributes new knowledge to the literature by exploring complexity
metrics and user comprehension of CMMN models. It attempted to theorically
and empirically validate the CMMN metrics. It contributed a novel approach to
the empirical validation of complexity metrics using pairwise comparisons. This
research also makes contributions in the areas of formalizing CMMN by using
first-order logic, identifying and comparing the model complexity of CMMN
against other process modeling notations, clarifying the relationship between
CMMN and GSM, and proposing complexity metrics for CMMN.
Bibliography
[1] L. C. Briand, S. Morasca, and V. R. Basili. “Property-based Software
Engineering Measurement”. In: IEEE Transactions on Software Engineering
22.1 (1996), pp. 68–86. doi: 10.1109/32.481535.
[2] M. Daneva, R. Heib, and A. W. Scheer. Benchmarking Business Process
Models. Research Report 136. Institute for Information Systems, University
of the Saarland, 1996.
[3] K. Kluza. “Measuring Complexity of Business Process Models Integrated
with Rules”. In: Artificial Intelligence and Soft Computing: 14th International
Conference, ICAISC 2015, Zakopane, Poland, June 14-18, 2015, Proceedings,
Part II. Ed. by L. Rutkowski et al. Springer International Publishing, 2015,
pp. 649–659. doi: 10.1007/978-3-319-19369-4_57.
[4] M. A. Marin. Research data, sources and documents for thesis on Exploring
Complexity Metrics for Artifact-Centric Business Process Models. Dataset
on Zenodo. 2018. doi: 10.5281/zenodo.1241595.
[5] A. Nigam and N. S. Caswell. “Business Artifacts: An Approach to Op-
erational Specification”. In: IBM Systems Journal 42.3 (2003), pp. 428–
445.
[6] OMG. Case Management Model and Notation, version 1.0. Standard. Docu-
ment formal/2014-05-05. OMG, 2014, p. 96.
[7] W. M. P. van der Aalst and P. J. S. Berens. “Beyond Workflow Management:
Product-Driven Case Handling”. In: Proceedings of the 2001 International
ACM SIGGROUP. New York: ACM Press, 2001, pp. 42–51. doi: 10.1145/
500286.500296.
[8] E. J. Weyuker. “Evaluating Software Complexity Measures”. In: IEEE
Transactions on Software Engineering 14.9 (1988), pp. 1357–1365. doi:
10.1109/32.6178.