Publications - Analysis/Mining

Publications in the area of Process Mining and Process Analysis


Mining the Organisational Perspective in Agile Business Processes
Stefan Schönig, Stefan Jablonski, Cristina Cabanillas, Jan Mendling @ 16th IFIP WG8.1 Working Conference on Business Process Modeling, Development, and Support (BPMDS 2015), in conjunction with CAiSE’15
Agile processes depend on human resources, decisions and expert knowledge, and they are especially versatile and comprise rather complex scenarios. Declarative, i.e., rule-based, process models are well suited for modelling these processes. Although there are several mining techniques to discover such declarative process models from event logs, they put less emphasis on the organisational perspective, which specifies how resources are involved in the activities. As a consequence, the resulting models do not specify who should execute which task and with which constraint (like separation of duties) in mind. In this paper, we propose a process mining approach to discover resource-aware, declarative process models. Our specific contribution is the extraction of complex rules for resource assignment that integrate the control- ow and organisational perspectives. Our experiments demonstrate the expressiveness of the mined rules with a reference to the Work flow Resource Patterns and a real-world use case.


Supporting Rule-based Process Mining by User-Guided Discovery of Resource-Aware Frequent Patterns
Stefan Schönig, Florian Gillitzer, Michael Zeising, Stefan Jablonski @ 1st Workshop on Resource Management in Service-Oriented Computing (RMSOC), in conjunction with ICSOC 2014
Agile processes depend on human resources, decisions and expert knowledge and are especially versatile and comprise rather complex coherencies. Rule-based process models are well-suited for modelling these processes. There exist a number of process mining approaches to discover rule-based process models from event logs. However, existing rule-based approaches are typically based on a given set of rule templates and predominately consider control flow aspects. By only considering a given set of templates, contemporary approaches underlie a representational bias. The usage of a fixed language frequently ends into insuffcient languages. In this paper we propose an approach to automatically suggest adequate resource-aware rule templates for a given domain by pre-processing the provided event log using frequent pattern mining techniques. These templates can then be instantiated and checked by process mining methods.


Supporting Collaborative Work by Learning Process Models and Patterns from Cases
Stefan Schönig, Michael Zeising, Stefan Jablonski @ 9th IEEE International Conference on Collaborative Computing: Networking, Applications and Worksharing (CollaborateCom 2013)
Recent work shows an increasing interest of the Busi-ness Process Management (BPM) community in unstructured, so-called "human-centric" processes. Case Management (CM) is a new trend that focuses on the support of collaborative human-cen-tric processes. Although CM provides concepts that support hu-man-centric work, processes have to be modelled beforehand in order to be supported by IT systems. Hence, a problem that arises when applying CM is that when organisations begin to formalize CM practice, it is often difficult to express rules controlling the ap-plicability of tasks. Furthermore, fundamental complexity chal-lenges arise when applying CM in practice. In this contribution, we provide a solution to these two issues. We propose that manag-ing human-centric processes should start with model skeletons that serve as a lattice where initial process execution can lean against. Additionally, by tracking different process cases, substan-tial process knowledge is recorded. Exploring process history might reveal certain recurring patterns that serve as dynamic guidance enhancement for CM systems. In this way, process mod-els might evolve over time, become more and more complete and better reflect operational reality.
Comprehensive Business Process Management through Observation and Navigation
Stefan Schönig, Michael Zeising, Stefan Jablonski @ 6th IFIP WG 8.1 working conference on the Practice of Enterprise Modeling (PoEM 2013)
Most real-world business processes involve a combination of both well-defined and previously modelled as well as unforeseen and therefor unmodelled scenarios. The goal of comprehensive process management should be to cover all actual-ly performed processes by accurate models so that they may be fully supported by IT systems. Unmodelled processes can be observed by the Process Observa-tion system which generates models reflecting the recorded behaviour. Modelled processes may be of different natures: while so-called “automation” processes in-volve little human participation and mainly orchestrate services and applications, so-called “knowledge-intensive” processes are based on human expert participa-tion. Both types of models may be enacted by the Process Navigation system. This contribution introduces the integration of both systems which leads to an approach for supporting the full range from unmodelled processes to both auto-mation and knowledge-intensive processes as well as the transition from unmod-elled to modelled processes.


Process Observation as Support for Evolutionary Process Engineering
Stefan Schönig, Michael Seitz, Michael Zeising, Stefan Jablonski @ International Journal On Advances in Systems and Measurements
The Process Observation project is used to generate business process execution logs and guides process participants through process execution. In this contribution, we introduce process evolution as an economic field of application for process observation. There are different needs for process evolution, e.g., to establish more consistent process results, continuously measure and improve process performance or meet accreditation requirements. We will show how process discovery, process guidance and process evidence as the main basic functions of process observation can be applied as support for reasonable process evolution. In this way, process observation can be used to reach a desired evolution stage or rather facilitate the transition between two maturity levels. Furthermore, process observation serves as an implementation for certain evolution stages itself and can additionally be consulted to prove the conformance to quality requirements of maturity levels.
Dynamic Guidance Enhancement in Workflow Management Systems
Christoph Günther, Stefan Schönig, Stefan Jablonski @ 27th Symposium On Applied Computing (SAC 2012)
Today’s workflow management systems have become increasingly powerful. Some prototypic approaches even tend to not patronise the users by providing a set of process steps to follow, but let them decide which step to choose next. The idea behind this approach is the impossibility to model every special case of a workflow, because a fixed process order would necessarily be inefficient or even incorrect in some cases. By admitting this freedom, the risk of confounding the users is taken. That is why we provide a qualified guidance instance through the process.
Discovering Cross-Perspective Semantic Definitions from Process Execution Logs
Stefan Schönig, Christoph Günther, Michael Zeising, Stefan Jablonski @ The Second International Conference on Business Intelligence and Technology (BUSTECH 2012)
In this paper, we suggest a two-phase declarative process mining approach discovering explicit, cross-perspective semantics. Cross-perspective semantics are interesting and important for the analysis of business processes, because they reveal dependencies that are not obvious on the first look. They allow for a comprehensive examination of the recorded process execution information and enable the discovery of coherency between different process-involved entities and perspectives. Using the described cross-perspective semantics, we additionally introduce an approach for simplifying less-structured process models.
Adapting Association Rule Mining to Discover Patterns of Collaboration in Process Logs
Stefan Schönig, Michael Zeising, Stefan Jablonski @ 8th IEEE International Conference on Collaborative Computing: Networking, Applications and Worksharing (CollaborateCom 2012)
The execution order of work steps within business processes is influenced by several factors, like the organizational position of performing agents, document flows or temporal dependencies. Lately, process mining techniques are more and more successfully used to discover execution orders from process execution logs automatically. Although, these techniques have been applied in various domains, the methods are mostly discovering the execution order of process steps without facing possible coherency with other perspectives of business processes, i.e., other types of process execution data. The reasons, e.g., for a given execution order, remain mostly undiscovered. In this paper, we propose a method to discover cross-perspective collaborative patterns in process logs and therefore strive for a genotypic anal-ysis of recorded process data. For this purpose, we adapted the association rule mining algorithm to analyse execution logs. The resulting rules can be used for guiding users through collaborative process execution.
Process Discovery and Guidance Applications of Manually Generated Logs
Stefan Schönig, Christoph Günther, Stefan Jablonski @ The Seventh International Conference on Internet Monitoring and Protection (ICIMP 2012)
In this paper, we investigate the problem of the availability of complete process execution event logs in order to offer automatic process model generation (process discovery) possibility by process mining techniques. Therefore, we present our AI4 | Process Observation Project that generates manual logs and guides process participants through process execution. Like this, our project offers the possibility for the automatic generation of process models within organizations, without the availability of any information system. Process participants are encouraged to work with our AI4 | Process Observation Tool by various process execution support functions, like an auto-suggestion of process data and dynamic recommendations of following processes.