Automated Planning of Process Models

?

In today’s increasingly complex business world, the capability to flexibly adapt existing or construct new processes is a vital skill for firms to stay competitive and to respond quickly. Business Process Management (BPM) and in particular approaches aiming to increase flexibility of processes are supporting stakeholders on these issues. Process modeling has proven to be a crucial instrument, for instance, for representing increasingly complex processes within and across firms, developing information systems and conducting business reorganization projects. However, it is often still time consuming and costly as it is usually realized in a manual manner even if it is supported by means of modeling tools or reference models.

With this project, we align to a research strand we named ‘Automated Process and Service Management’ which aims to increase process flexibility by means of automation, allowing to support business analyst and modelers in different phases of the BPM Lifecycle. For instance, process mining assists business analysts in the process analysis phase. Automated service selection increases the degree of automation within the phases process implementation as well as process execution and process planning supports modelers by constructing process models automatically in the process modeling phase. This project aimed at developing theoretical foundations, models, and in particular algorithms for an automated construction and adaption of semantically annotated process models. In the following, we use the term automated planning of process models to refer to this. Additionally, we aimed at developing approaches that allow for an automated and optimization-based selection of services. We thereby focused at least on both quality of service (QoS)-criteria (e.g., price) as economic performance measures and resource restrictions (e.g., an upper limit regarding the price of a single process execution). In the following, we use the term QoS-aware service selection to refer to this. With respect to the phases of the BPM lifecycle, the automated planning of process models addresses the phase process modeling and the QoS-aware service selection concentrates on the phases process implementation and process execution (cf. Figure 1).

?

Figure1: BPM Lifecycle

?

Particularly, we dealt with the following main goals:

[G1]: Automated construction of process models considering control flow structures and multiple process goals

[G2]: Automated adaption of existing process models

[G3]: Automated QoS-aware service selection considering economic performance measures and resource restrictions

[G4]: Evaluation

?

Figure2: Schematic overview of the developed approaches

?

Most important results and brief description of their significance (main points):

In the following, we will discuss the main results with respect to the goals [G1] to [G4]:

[G1] ?? Automated construction of process models considering control flow structures and multiple process goals

With this goal, we aimed at developing approaches that are able to construct control flow structures (cf. step 3 in Figure 4) in an automated manner that are used in process models. Both the basic concept and the foundation for this is published in Heinrich et al. (2011a) and Heinrich et al. (2008). The research paper on the automated planning of exclusive choices has been published in Decision Support Systems (Heinrich et al. 2015b). The complete research paper on the automated construction of simple merges has been accepted for publication and presentation at the European Conference on Information Systems (ECIS) 2016 (Heinrich and Sch?n 2016). The paper on the automated planning of parallel splits and synchronizations can be resubmitted to Information Systems Research (Heinrich et al. 2016d) after a revision, which is currently done.

Besides the automated planning of control flow structures, the requirement of considering multiple process goals has to be addressed in [G1]. This requirement is especially relevant because processes nowadays usually span across different departments or even firms (i.e., actors). To support the automated planning of such processes, we developed a theoretical basis and algorithms that allow coordinating multiple participating actors to enable each actor to reach his/her individual goal. Current approaches for an automated planning do not aim for such a support. The corresponding complete research paper on the coordination of multiple actors by means of automated planning is finished and has been submitted to International Conference on Information Systems (ICIS) 2016 (Heinrich et al. 2016b).

[G2] Automated adaption of existing process models

Many existing process models need to be adapted frequently due to, for instance, new customer demands, market shifts or new regulatory laws. We therefore envisioned the automated adaption of existing process models with goal [G2] and developed a corresponding approach. The research paper presenting this approach has been submitted to ICIS 2016 (Heinrich et al. 2016a). Furthermore, most real world processes are heavily influenced by environmental factors such as context information. Here, existing approaches consider context rather as a static information although it is known to be highly dynamic. To offer support for processes that are influenced by changing context information, we developed both a modelling concept and an algorithm to cope with the dynamic nature of context information in process models. We therefore enabled an automated adaption of process models with respect to changing context information. The corresponding complete research paper has been accepted and presented at ECIS 2015 (Heinrich and Sch?n 2015).

[G3] ?? Automated QoS-aware service selection considering economic performance measures and resource restrictions

Regarding goal [G3], we analyzed how the degree of automation in the BPM lifecycle could be further increased by approaches that allow for an automated selection of services (e.g., web services & mobile services) while considering economic performance measures, resource restrictions, and external and internal influencing factors (e.g., context information and multiple actors) – see step 5 in Figure 4. To address this goal, we developed two approaches to assess and select web services based upon QoS-criteria (i.e., costs, response time, etc.) while taking potential service failures and nondeterministic service values into account. The first complete research paper has been accepted and presented at the International Conference on Wirtschaftsinformatik 2013 (Heinrich and Lewerenz 2013). This paper was significantly enhanced by a second paper, which means, its scope was broadened by the consideration of nondeterministic service values as well as different control flow structures, it was mathematically re-defined and the evaluation was considerably extended. The second paper has been published in the Informs Journal Service Science (Heinrich et al. 2015c). Moreover, we developed approaches to determine the optimal granularity of services for process implementation. Here, the first article has been published in Business & Information Systems Engineering (Heinrich et al. 2011b). The second complete research paper has been accepted and presented at ICIS 2012 (Heinrich and Zimmermann 2012). Concerning the efficient selection of services, we developed three approaches to consider context information and multiple actors. This allows selecting services (especially services in a mobile context) that are mostly tailored to the current needs and situations of a single actor or of multiple actors. The first approach has been published in the Journal of Decision Systems (Heinrich and Lewerenz 2015a). Based upon this foundation, a second approach was developed that particularly focuses on the computational complexity when selecting (mobile) services. This complete research paper has been accepted and presented at ICIS 2015 (Lewerenz 2015). Further, a complete research paper has been accepted and presented at ICIS 2015 (Heinrich et al. 2015b) that deal with multiple actors and thus allows to take different types of dependencies (e.g., of temporal nature) among participating actors into account when selecting (mobile) services.

[G4] Evaluation

We addressed this goal in each of the previously mentioned papers as we evaluated each of the developed approaches – in a first step – on its own. For this purpose, we provided mathematical proofs regarding the completeness, minimality, termination, computational complexity, etc. of the proposed approaches for the goals [G1] to [G2] (see Heinrich and Sch?n 2015, Heinrich et al. 2015b, Heinrich et al. 2015d, Heinrich and Sch?n 2016, Heinrich et al. 2016a, Heinrich et al. 2016b). We also used mathematical proofs to illustrate the benefits of the approaches for goal [G3] (see Heinrich and Lewerenz 2013, Heinrich et al. 2015c). In addition to that, we also conducted several extensive simulation experiments to demonstrate especially the applicability and efficacy of the approaches for goal [G1] to [G3] (see Heinrich et al. 2015c, Heinrich et al. 2016a, Heinrich and Lewerenz 2015a, Lewerenz 2015). Finally, we operationalized the developed approaches by means of prototypical implementations and applied them to several real-world scenarios. For instance, in a cooperation with the HypoVereinsbank AG, we constructed numerous process models with the help of our approaches (see, e.g., Heinrich et al. 2015b). For the research on coordinating multiple actors, we applied our algorithm to a real-world manufacturing process (Heinrich et al. 2016b). In addition to that, we prototypically implemented the developed approaches as a software-as-a-service and developed a web-based graphical modeling tool (cf., http://www-sempa.uni-regensburg.de; see Figure 3). We thereby conducted several extensive integration tests to guarantee for correct results across all developed approaches and carefully considered the computational complexity of the overall approach for the automated planning of process models.

?

Figure 3: Screenshot of the web-based graphical modeling tool

?

Further, we conducted a laboratory experiment on the influence of automated process planning on the efficiency of process modelers. As part of different lectures at the University of Regensburg, students had to construct process models either manually or supported by the approaches developed in this research project. We observed that using the approaches significantly increases the process modeler efficiency. Especially the completeness and correctness of the constructed process models could be improved, where at the same time the required personnel efforts could be decreased. A complete research paper summarizing the detailed results of this experiment has been submitted to ICIS 2016 (Sch?n 2016).

Figure 4: Procedural overview of automated planning of process models and QoS-aware service selection

?

Peer-reviewed publications / already published

Heinrich, B., and Sch?n, D. (2016). Automated Planning of Process Models: The Construction of Simple Merges. In: Proceedings of the 24th European Conference on Information Systems (ECIS), Istanbul, Turkey. (Green OA; http://epub.uni-regensburg.de/33576/)

Lewerenz, L. (2015). A Heuristic Technique for an Efficient Decision Support in Context-aware Service Selection. In: Proceedings of the 36th International Conference on Information Systems (ICIS), Fort Worth, USA. (Green OA; http://epub.uni-regensburg.de/32660/)

Heinrich, B., Klier, M., Lewerenz, L., and Mayer, M. (2015a). Enhancing Decision Support in Multi User Service Selection. In: Proceedings of the 36th International Conference on Information Systems (ICIS), 2015, Fort Worth, USA. (Green OA; http://epub.uni-regensburg.de/32558/)

Heinrich, B., and Lewerenz, L. (2015a). Decision Support for the Usage of Mobile Information Services: A Context-aware Service Selection Approach that Considers the Effects of Context Interdependencies. Journal of Decision Systems (24:4), pp. 406-432. (Green OA; http://epub.uni-regensburg.de/32639/)

Heinrich, B., Klier, M., and Zimmermann, S. (2015b). Automated Planning of Process Models: Design of a Novel Approach to Construct Exclusive Choices. Decision Support Systems (78), October 2015, pp. 1–14. (Green OA; http://epub.uni-regensburg.de/32342/)

Heinrich, B., and Sch?n, D. (2015). Automated Planning of Context-aware Process Models. In: Proceedings of the 23rd European Conference on Information Systems (ECIS), Münster, Germany. (Green OA; http://epub.uni-regensburg.de/31454/)

Heinrich, B., Klier, M., Lewerenz, L., and Zimmermann, S. (2015c). Quality of Service-aware Service Selection: A Novel Approach Considering Potential Service Failures and Nondeterministic Service Values. INFORMS Service Science (7:1), March 2015, pp. 1-22. (Green OA; http://epub.uni-regensburg.de/31231/)

Heinrich, B., and Lewerenz, L. (2013). QoS-aware Service Selection Considering Potential Service Failures. In: Proceedings of the 11th International Conference Wirtschaftsinformatik, Leipzig, Germany. (Green OA; http://epub.uni-regensburg.de/27168/)

Heinrich, B., and Zimmermann, S. (2012). Granularity Metrics for IT Services. In: Proceedings of the 33rd International Conference on Information Systems (ICIS), Orlando/Florida, USA. (Green OA; http://epub.uni-regensburg.de/27167/)

Heinrich, B., Klier, M., and Zimmermann, S. (2011a). Automatisierte Modellierung, Umsetzung und Ausführung von Prozessen – Ein Web Service-basiertes Konzept. In: Proceedings of the 10th International Conference Wirtschaftsinformatik, Zürich, Schweiz. (Green OA; http://epub.uni-regensburg.de/27167/)

Heinrich, B., Henneberger, M., Krammer, A., and Lautenbacher, F. (2011b). Granularity of Serivces?–?an Economic Analysis. In: Business & Information Systems Engineering (3:6), pp. 345?–?358 (Green OA; http://epub.uni-regensburg.de/23814/)

Heinrich, B., Bewernik, M.A., Henneberger, M., Krammer, A., and Lautenbacher. (2008). SEMPA?–?A Semantic Business Process Management Approach for the Planning of Process Models. WIRTSCHAFTSINFORMATIK (50:6), pp. 445?–?460. (in German)