Joachim Herbst

A machine learning approach to workflow management

There has recently been some interest in applying machine learning techniques to support the acquisition and adaptation of workflow models. The different learning algorithms, that have been proposed, share some restrictions, which may prevent them from being used in practice. Approaches applying techniques from grammatical inference are restricted to sequential workflows. Other algorithms allowing concurrency require unique activity nodes. This contribution shows how the basic principle of our previous approach to sequential workflow induction can be generalized, so that it is able to deal with concurrency. It does not require unique activity nodes. The presented approach uses a log-likelihood guided search in the space of workflow models, that starts with a most general workflow model containing unique activity nodes. Two split operators are available for specialization.

Integrating machine learning and workflow management to support acquisition and adaptation of workflow models

Current workflow management systems (WFMS) offer little aid for the acquisition of workflow models and their adaptation to changing requirements. To support these activities we propose to integrate machine learning and workflow management. This enables an inductive approach to workflow acquisition and adaptation by processing traces of manually enacted workflows. We present a machine learning component that combines two different machine learning algorithms. In this paper we focus mainly on the first one, which induces the structure of the workflow, based on the induction of hidden markov models. The second algorithm, a standard decision rule induction algorithm, induces transition conditions. The main concepts have been implemented in a prototype, which we have validated using artificial process traces. The induced workflow models can be imported by the business process management system ADONIS.

Data-driven modeling and coordination of large process structures

In the engineering domain, the development of complex products (e.g., cars) necessitates the coordination of thousands of (sub-) processes. One of the biggest challenges for process management systems is to support the modeling, monitoring and maintenance of the many interdependencies between these sub-processes. The resulting process structures are large and can be characterized by a strong relationship with the assembly of the product; i.e., the sub-processes to be coordinated can be related to the different product components. So far, subprocess coordination has been mainly accomplished manually, resulting in high efforts and inconsistencies. IT support is required to utilize the information about the product and its structure for deriving, coordinating and maintaining such data-driven process structures. In this paper, we introduce the COREPRO framework for the data-driven modeling of large process structures. The approach reduces modeling efforts significantly and provides mechanisms for maintaining data-driven process structures.

A New Paradigm for the Enactment and Dynamic Adaptation of Data-Driven Process Structures

Industry is increasingly demanding IT support for large engineering processes, i.e., process structures consisting of hundreds up to thousands of processes. Developing a car, for example, requires the coordination of development processes for hundreds of components. Each of these development processes itself comprises a number of interdependent processes for designing, testing, and releasing the respective component. Typically, the resulting process structure becomes very large and is characterized by a strong relation with the assembly of the product. Such process structures are denoted as data-driven. On the one hand, the strong linkage between data and processes can be utilized for automatically creating process structures. On the other hand, it is useful for (dynamically) adapting process structures at a high level of abstraction. This paper presents new techniques for (dynamically) adapting data-driven process structures. We discuss fundamental correctness criteria needed for (automatically) detecting and disallowing dynamic changes which would lead to an inconsistent runtime situation. Altogether, our COREPRO approach provides a new paradigm for changing data-driven process structures at runtime reducing costs of change significantly.

IT support for release management processes in the automotive industry

Car development is based on long running, concurrently executed and highly dependent processes. The coordination and synchronization of these processes has become a complex and error-prone task due to the increasing number of functions and embedded systems in modern cars. These systems realize advanced features by embedded software and enable the distribution of functionality as required, for example, by safety equipment. Different life cycle times of mechanical, software and hardware components as well as different duration of their development processes require efficient coordination. Furthermore, product-driven process structures, dynamic adaptation of these structures, and handling real-world exceptions result in challenging demands for any IT system. In this paper we elaborate fundamental requirements for the IT support of car development processes, taking release management as characteristic example. We show to which extent current product data and process management technology meets these requirements, and discuss which essential limitations still exist. This results in a number of fundamental challenges requiring new paradigms for the product-driven design, enactment and adaptation of processes.

Workflow mining with InWoLvE

State of the art information systems are based on explicit process models called workflow models. Experience from industrial practice shows that the definition of workflow models is a very time consuming and error prone task. Recently, there has been an increasing interest in applying techniques from data mining and machine learning to support this task. This approach has also been termed as process or workflow mining. In this paper, we give an overview of the algorithms that were implemented within the InWoLvE workflow mining system, we summarize the most important results of their experimental evalualion and we present the experiences that were made in the first industrial application of InWoLvE.

Flexibility of data-driven process structures

The coordination of complex process structures is a fundamental task for enterprises, such as in the automotive industry. Usually, such process structures consist of several (sub-)processes whose execution must be coordinated and synchronized. Effecting this manually is both ineffective and error-prone. However, we can benefit from the fact that these processes are correlated with product structures in many application domains, such as product engineering. Specifically, we can utilize the assembly of a complex real object, such as a car consisting of different mechanical, electrical or electronic subcomponents. Each sub-component has related design or testing processes, which have to be executed within an overall process structure according to the product structure. Our goal is to enable product-driven (i.e., data-driven) process modeling, execution and adaptation. We show the necessity of considering the product life cycle and the role of processes, which are triggering state transitions within the product life cycle. This paper discusses important issues related to the design, enactment and change of data-driven process structures. Our considerations are based on several case studies we conducted for engineering processes in the automotive industry.

Interactive workflow mining: requirements, concepts and implementation

Many information systems log event data about executed tasks. Workflow mining is concerned with the derivation of a graphical workflow model out of this data. Experience from applying our workflow mining system InWoLvE in experiments and practical applications has shown that workflow mining is a highly interactive process. The mining expert iteratively approaches the result by varying the parameters of the mining tool and verifying the mined models. Our tool InWoLvE was not designed for intensive interactive usage. In this paper, we report about a rigorous requirements analysis and about possible solutions related with the support of such interactivity. Two selected solution concepts are explained in more detail. First, a special layout algorithm that is stable against small changes of the model thus allowing the workflow mining expert to maintain a mental map of the workflow. Second, a validation procedure that helps the expert to check event sequences against the (preliminary) mined model. These and other important concepts have been implemented in a prototype called ProTo.

COREPROSim: A Tool for Modeling, Simulating and Adapting Data-Driven Process Structures

Industry is increasingly demanding IT support for large engineering process structures consisting of hundreds up to thousands of synchronized processes. In technical domains, such process structures are characterized by their strong relation to the assembly of a product (e.g., a car); i.e., resulting process structures are data-driven. The strong linkage between data and processes can be utilized for automatically creating process structures as well as for (dynamically) adapting them at a high level of abstraction. This paper presents the COREPRO Sim demonstrator which enables sophisticated support for modeling, coordinating and (dynamically) adapting data-driven process structures. COREPRO Sim substantiates the COREPRO approach which provides a new paradigm for the integration of complex data and process structures.

Interactive workflow mining

Workflow or process mining is concerned with deriving a workflow model from observed behavior described in a workflow log. Experience from applying our workflow mining system InWoLvE in experiments and practical applications has shown that workflow mining is a highly interactive process. The mining expert iteratively approaches the result by varying the parameters of the mining tool and verifying the mined models. Our tool InWoLvE was not designed for intensive interactive usage making practical usage more than difficult. In this contribution we describe the main requirements for an interactive workflow mining system and how we derived these. We outline two selected concepts: a special layout algorithm that is stable against small changes of the model thus allowing the workflow mining expert to maintain a mental map of the workflow and a validation procedure helping the mining expert in his decision for the final result. These and other important concepts have been implemented in the first prototype of an interactive workflow mining system called ProTo.