These systems include not only standard Workflow management systems (WFMS), which execute and control processes explicitly, but also systems such as ERP (Enterprise Resource Planning) systems, CRM (Customer Relationship Management) systems, high-end middleware (WebSphere), etc.; all these systems imply that there is a process notion and that they are aware of the supported processes. Since real processes should be reliable and effective and should match corresponding process models, PAIS require tools and methods to automatically control execution of real processes within a system.
Goal of Research: The goal of the research is to develop new and extend existing approaches for modeling, analysis, improvement and monitoring processes within PAIS. These approaches include Process mining techniques (methods for discovering process models from event logs, checking process models against event logs) and methods for constructing advanced formal models for process analysis.
Empirical Base of Research: Real life event logs from municipalities and healthcare were employed to verify new, and extend existing Process mining methods. Unique methods and approaches were implemented within the ProM (Process mining framework).
Results of Research:
Further theoretical results were obtained in the field of modeling and analysis of complex information systems.
The possibility of translating Nested Petri nets (NP-nets) into Colored Petri nets (CPN) was proven and used for simulation and verification of NP-net models of complex distributed systems. An extended data scheme, a graphical editor and a code generator for NP-nets were all developed.
A novel formalism of Timed Resource Driven Automata Nets (TRDA-nets) and a tool for modelling and analysis of complex distributed real-time systems of mobile agents with the help of this formalism were developed. A formalism of Resource Driven Automata Nets allows us to visually express concurrency in terms of autonomous (asynchronous) behavior, spatial distribution of objects (agents/resources) and sharing access to common resources.
The service compatibility problem is to answer the question as to whether two Web services fit together, i.e. whether the composed system is sound. The algorithm of checking necessary conditions for the service compatibility problem was developed and proved to be correct and efficient.
Existing approaches for applying Process mining techniques to services (service mining) and inter-organizational workflows were analyzed and reviewed. Based on this analysis, further research directions were formulated; they form a background for developing innovative service mining techniques and analyzing the inter-organizational workflows.
In the field of Process mining the following theoretical results were obtained.
A novel approach based on the notion of process cubes was introduced. In process cubes, events and process models are organized using different dimensions. Each cell in the process cube corresponds to a set of events and can be used to discover a process model, to check conformance with respect to some process model, or to discover bottlenecks. The idea is related to the well-known OLAP (Online Analytical Processing) data cubes and associated operations such as slice, dice, roll-up, and drill-down. The notion of process cubes was formalized.
A novel generic approach for log and model decomposition was proposed. Most Process mining algorithms are linear in the size of the event log and exponential in the number of different activities. Therefore, it is reasonable to decompose large process mining problems into collections of smaller process mining problems focusing on restricted sets of activities. For conformance checking, the process model is decomposed into smaller partly overlapping sub-models using projection. The event log is decomposed into sub-logs, also using projection. Any trace that ﬁts into the overall model also ﬁts all sub-models. This work gives the background for decomposing process mining problems.
A decomposition method for model repair was developed.
A method for discovering readable process models from event logs with cancellations was developed. It was shown that occurrence of cancellations in a log frequently leads to process models with an overcomplicated control flow. A novel algorithm for discovering cancellations and constructing a RWF-net (Reset workflow-net) with a more compact and transparent structure was presented. The correctness of this algorithm was proven.
Moreover, the following applied results in Process mining area were obtained.
The DPMine tool allows the visualization and execution of log analysis steps. Individual units of work are connected to each other within a process graph. A language was developed to formalize these processes graphs.
ProM plugins support plenty of different process model formats. Such commonly known and widely accepted process modeling standard as Business Process Modeling and Notation (BPMN) was integrated into ProM. Import and export capabilities of ProM were extended to support integration with existing BPMN modeling tools (such as Signavio and Bisagi). Additionally, ProM plugins for conversion from (Data) Petri nets, Casual nets and Process trees to BPMN were developed.
A program component for loading logs into a database was developed. Memory usage and network load parameters were analyzed for both cases when the log is stored in a database or saved as a file.
Methods of storing multidimensional data for log analysis were proposed. A system for loading log data into a multidimensional warehouse and a user access system were developed.
Theoretical and practical results were also obtained in the theory of Colored Petri nets: CPN Tools 4 with new capabilities was released. A tool called Grade/CPN which supports the grading of Colored Petri nets modeled in CPN Tools was developed. A novel temporal logic (BTL) and a new approach for modelling workflows combining the procedural formalism colored Petri nets, and the two declarative formalisms, Declare and DCR (Dynamic Condition Response Graphs) graphs were presented. The Unconstrained Miner, a tool for fast and accurate mining Declare constraints from models without imposing any assumptions about the model, was developed. Also a review of existing four workflow paradigms was done. This review helps to consider a model bias in a broader sense.
The applied research done in PAIS lab has been directed to investigate applicability of Process mining techniques to typical real cases. For this reason healthcare was investigated. Analysis of concrete case studies was performed using process mining techniques and ProM framework. These case studies provided many interesting insights. It was shown, that process mining is applicable and helpful for finding patterns in real processes.
Level of implementation, recommendations on implementation or outcomes of the implementation of Results: Thetheoretical results presented in this work were implemented and verified within various platforms such as ProM framework, DPMine, Tool for modelling TRDA-nets, CPN Tools, Declare, Unconstrained Miner. Most of the theoretical results gave a mature background for further research. The applied research in its turn provided interesting insights of the real-life processes, the results of this research can be effectively used for further fundamental studies.
Field of application: The approaches developed during the research help to model, analyze, improve and monitor real-life processes within PAIS in various areas, such as healthcare, booking systems, government and others. This is crucial for reducing the costs and optimizing existing business processes and also for improving the quality of the information systems in general.