Dominic Breuker

Comprehensible predictive models for business processes

Predictive modeling approaches in business process management provide a way to streamline operational business processes. For instance, they can warn decision makers about undesirable events that are likely to happen in the future, giving the decision maker an opportunity to intervene. The topic is gaining momentum in process mining, a field of research that has traditionally developed tools to discover business process models from data sets of past process behavior. Predictive modeling techniques are built on top of process-discovery algorithms. As these algorithms describe business process behavior using models of formal languages (e.g., Petri nets), strong language biases are necessary in order to generate models with the limited amounts of data included in the data set. Naturally, corresponding predictive modeling techniques reflect these biases. Based on theory from grammatical inference, a field of research that is concerned with inducing language models, we design a new predictive modeling technique based on weaker biases. Fitting a probabilistic model to a data set of past behavior makes it possible to predict how currently running process instances will behave in the future. To clarify how this technique works and to facilitate its adoption, we also design a way to visualize the probabilistic models. We assess the effectiveness of the technique in an experimental evaluation with synthetic and real-world data.

Can We Afford Integrity by Proof-of-Work? Scenarios Inspired by the Bitcoin Currency

Proof-of-Work (PoW), a well-known principle to ration resource access in client-server relations, is about to experience a renaissance as a mechanism to protect the integrity of a global state in distributed transaction systems under decentralized control. Most prominently, the Bitcoin cryptographic currency protocol leverages PoW to (1) prevent double spending and (2) establish scarcity, two essential properties of any electronic currency. This chapter asks the important question whether this approach is generally viable. Citing actual data, it provides a first cut of an answer by estimating the resource requirements, in terms of operating cost and ecological footprint, of a suitably dimensioned PoW infrastructure and comparing them to three attack scenarios. The analysis is inspired by Bitcoin, but generalizes to potential successors, which fix Bitcoin’s technical and economic teething troubles discussed in the literature.

Towards Risk Scoring of Bitcoin Transactions

If Bitcoin becomes the prevalent payment system on the Internet, crime fighters will join forces with regulators and enforce blacklisting of transaction prefixes at the parties who offer real products and services in exchange for bitcoin. Blacklisted bitcoins will be hard to spend and therefore less liquid and less valuable. This requires every recipient of Bitcoin payments not only to check all incoming transactions for possible blacklistings, but also to assess the risk of a transaction being blacklisted in the future. We elaborate this scenario, specify a risk model, devise a prediction approach using public knowledge, and present preliminary results using data from selected known thefts. We discuss the implications on markets where bitcoins are traded and critically revisit Bitcoin’s ability to serve as a unit of account.

Evaluating the Expressiveness of Domain Specific Modeling Languages Using the Bunge-Wand-Weber Ontology

Business Process Management is becoming an ever more important aspect for organizations alongside with Business Process Diagrams as a tool to describe business processes. So far process modeling has been mainly performed with generic process modeling languages. These approaches have however limitations when it comes to the needs of specific problem domains or automated process analysis. Semantic building block based languages (SBBL) aim to overcome those limitations by integrating domain semantics in the modeling language. However, this class of languages is only useful if they exhibit the same expressiveness as generic languages. In this paper we strive to answer this question by comparing the expressiveness of the SBBL language PICTURE with ARIS as a generic language based on the Bunge-Wand-Weber ontology, showing that PICTURE has hardly construct deficits compared to ARIS while showing less construct redundancy and construct overload in its constructs.

Towards Model-Driven Engineering for Big Data Analytics -- An Exploratory Analysis of Domain-Specific Languages for Machine Learning

Graphical models and general purpose inference algorithms are powerful tools for moving from imperative towards declarative specification of machine learning problems. Although graphical models define the principle information necessary to adapt inference algorithms to specific probabilistic models, entirely model-driven development is not yet possible. However, generating executable code from graphical models could have several advantages. It could reduce the skills necessary to implement probabilistic models and may speed up development processes. Both advantages address pressing industry needs. They come along with increased supply of data scientist labor, the demand of which cannot be fulfilled at the moment. To explore the opportunities of model-driven big data analytics, I review the main modeling languages used in machine learning as well as inference algorithms and corresponding software implementations. Gaps hampering direct code generation from graphical models are identified and closed by proposing an initial conceptualization of a domain-specific modeling language.

A Runtime Analysis of Graph-Theoretical Algorithms to Detect Patterns in Process Model Collections

Pattern detection serves different purposes in managing large collections of process models, ranging from syntax checking to compliance validation. This paper presents a runtime analysis of four graph-theoretical algorithms for (frequent) pattern detection. We apply these algorithms to large collections of process and data models to demonstrate that, despite their theoretical intractability, they are able to return results within (milli-) seconds. We discuss the relative performance of these algorithms and their applicability in practice.

Graph theory and model collection management: conceptual framework and runtime analysis of selected graph algorithms

Abstract Analysing conceptual models is a frequent task of business process management (BPM), for instance to support comparison or integration of business processes, to check business processes for compliance or weaknesses, or to tailor conceptual models for different audiences. As recently, many companies have started to maintain large model collections and analysing such collections manually may be laborious, practitioners have articulated a demand for automatic model analysis support. Hence, BPM scholars have proposed a plethora of different model analysis techniques. As virtually any conceptual model can be interpreted as a mathematical graph and model analysis techniques often include some kind of graph problem, in this paper, we introduce a graph algorithm based model analysis framework that can be accessed by specialized model analysis techniques. To prove that basic graph algorithms are feasible to support such a framework, we conduct a performance analysis of selected graph algorithms.

Designing and Evaluating an Interpretable Predictive Modeling Technique for Business Processes

Process mining is a field traditionally concerned with retrospective analysis of event logs, yet interest in applying it online to running process instances is increasing. In this paper, we design a predictive modeling technique that can be used to quantify probabilities of how a running process instance will behave based on the events that have been observed so far. To this end, we study the field of grammatical inference and identify suitable probabilistic modeling techniques for event log data. After tailoring one of these techniques to the domain of business process management, we derive a learning algorithm. By combining our predictive model with an established process discovery technique, we are able to visualize the significant parts of predictive models in form of Petri nets. A preliminary evaluation demonstrates the effectiveness of our approach.

Developing Graphical Model Editors for Meta-Modelling Tools - Requirements, Conceptualisation, and Implementation

Meta-modelling tools have been proposed to facilitate the development and adoption of domain-specific modelling languages (DSMLs). These languages specify a set of domain-specific concepts and assign diagrammatic representations to them. A considerable amount of work has been done to develop metamodelling tools ensuring syntactical correctness of models created with DSMLs. However, little has been published about the challenges of developing a graphical model editor for meta-modelling tools. Specifying how conceptual elements of a DSML are to be represented graphically is often cumbersome. Moreover, tools are sometimes too inflexible to handle advanced features beyond displaying static icons. Furthermore, graphical representations must be kept consistent in case of reuse in multiple, potentially integrated DSMLs. This paper’s aim is to carve out a set of requirements for graphical model editors as used in meta-modelling tools. We present a conceptual model considering these requirements. Furthermore, we discuss an exemplary software implementation of a model editor.

An Efficient Business Process Compliance Checking Approach

Assuring compliant business processes is an important task of business process management, which is commonly supported by the use of business process models. As every compliance rule corresponds with a typical structure, the detection of those corresponds to a pattern matching problem. More specifically, we encounter the problem of subgraph isomorphism. In this paper we propose an automatic business process compliance checking approach that relies on a subgraph isomorphism algorithm and that is suitable for process models in general. As common subgraph isomorphism is a problem that can only be solved in exponential time, we use an algorithm that simplifies the problem through pre-processing. This makes the isomorphism solvable in polynomial time. With the approach, we aim at supporting decision makers in business process compliance management.