Advanced Clustering Techniques for Business Process Execution Traces

In today's complex economic environment, organisations are interacting with an array of different actors such as customers, suppliers, competitors, governmental bodies, and other internal and external stakeholders. These interactions are comprised of a wide variety of business processes. The data stored in information systems supporting these processes contain valuable insights for these organisations. These business process execution traces are the focal point of this thesis text. A wide array of techniques exists for interpreting such process data, for example techniques for visualising stored behaviour, or techniques for checking whether past behaviour confirms to the normal way of working. This thesis specifically concentrates on approaches used for dealing with a highly varied set of business process execution traces, by splitting them into more meaningful and useful groups, called clustering. In a first part of the thesis, the fields of business process management, process mining and trace clustering are described. The content and contributions of this work are detailed as well. The main part of the thesis described advances to the previously described field of trace clustering. First, the usability of trace clustering techniques is enhanced by providing approaches for determining an appropriate number of clusters. Secondly, a technique is proposed for extracting explanations from clustering solutions, improving their interpretability. Finally, a number of new trace clustering techniques are proposed which incorporate expert knowledge to make trace clustering solutions more justifiable. In the subsequent part, two contributions are made to the broader field of process mining. On the one hand, a framework is proposed for evaluating the stability of process discovery techniques. Such techniques are used to extract visual process models from raw process data. On the other hand, a number of approaches is presented for learning representations of a number of business process artifacts, such as event logs and process models. The final part provides an overview of the contributions, and describes a number of avenues for future research.status: publishe