Matching of events and activities

Abstract

I like things matching. I have an upright bass, a drum kit and a grand piano that's the same color.-Penn Jillette A B S T R A C T Nowadays, business processes are increasingly supported by IT services that produce massive amounts of event data during process execution. Aiming at a better process understanding and improvement, this event data can be used to analyze processes using process mining techniques. Process models can be automatically discovered and the execution can be checked for conformance to specified behavior. Moreover, existing process models can be enhanced and annotated with valuable information, for example for performance analysis. While the maturity of process mining algorithms is increasing and more tools are entering the market, process mining projects still face the problem of different levels of abstraction when comparing events with modeled business activities. Mapping the recorded events to activities of a given process model is essential for conformance checking, annotation and understanding of process discovery results. Current approaches try to abstract from events in an automated way that does not capture the required domain knowledge to fit business activities. Such techniques can be a good way to quickly reduce complexity in process discovery. Yet, they fail to enable techniques like conformance checking or model annotation, and potentially create misleading process discovery results by not using the known business terminology.In this thesis, we develop approaches that abstract an event log to the same level that is needed by the business. Typically, this abstraction level is defined by a given process model. Thus, the goal of this thesis is to match events from an event log to activities in a given process model. To accomplish this goal, behavioral and linguistic aspects of process models and event logs as well as domain knowledge captured in existing process documentation are taken into account to build semiautomatic matching approaches. The approaches establish a pre-processing for every available process mining technique that produces or annotates a process model, thereby reducing the manual effort for process analysts. While each of the presented approaches can be used in isolation, we also introduce a general framework for the integration of different matching approaches.The approaches have been evaluated in case studies with industry and using a large industry process model collection and simulated event logs. The evaluation demonstrates the effectiveness and efficiency of the approaches and their robustness towards nonconforming execution logs. • Thomas Baier, Andreas Rogge-Solti, Mathias Weske, and Jan Mendling. Matching of events and activities -an approach based on behavioral constraint satisfaction. In Proceedings of the 30th Annual ACM Symposium on Applied Computing, pages 1225-1230. ACM, 2015.• Thomas Baier, Claudio Di Ciccio, Jan Mendling, and Mathias Weske. Matching of events and activities -an approach using declarative modeling constraints. Other important roles i...

From Low-Level Events to Activities - A Pattern-Based Approach

“…Generally, there can be an n:m-relation between recorded events and activities [2,3], i.e., one higher level activity may create multiple low level events and one such event possibly relates to multiple activities. There are proposals for unsupervised abstraction methods that try to determine this relation based on identifying sub-sequences and machine learning methods [2,4,5,6,7], as well as proposals for supervised methods based on existing process documentation and constraint satisfaction [3,8,9,10,11]. Unsupervised abstraction methods, clearly, do not take existing knowledge into account and may fail to provide meaningful labels for discovered event clusters.…”

“…Unsupervised abstraction methods, clearly, do not take existing knowledge into account and may fail to provide meaningful labels for discovered event clusters. Existing supervised abstraction methods [3,8,9,10,11] assume knowledge about a single model for the overall process. They resolve to clustering methods and heuristics when challenged with event logs from processes that feature n:m event-activity relations, concurrent activities, and noise (i.e., erroneous or missing events).…”

An Integrated Approach for Discovering Process Models According to Business Process Types

“…The methods developed by (Baier et. al., 2015;Baier, 2015;Baier et al, 2014) assume the knowledge of a single highlevel model for the overall process.…”

“…The methods developed by (Baier et. al., 2015;Baier, 2015;Baier et al, 2014) assume the knowledge of a single highlevel model for the overall process. The goal is to automatically discover the relation between events and activities.…”

Guided Process Discovery – A pattern-based approach

“…The method is limited to traces of length less than 30 events due to the computational complexity. Most related to our work are the methods developed by Baier et al [3,42,43]. Again, the methods assume knowledge about a single high-level model for the overall process.…”