Detecting Surprising Instances in a Process


Student: Christian Kohlschmidt

Title: Detecting Surprising Instances in a Process

Supervisor: Mahnaz Qafari

1st Examiner: Prof. Wil M.P. van der Aalst

2nd Examiner: Prof. Martin Grohe


Process mining is a set of techniques organizations use to understand and improve their operational processes. The first essential step to enhance a process is to discover the process improvement opportunities in the existing process. Existing approaches assume that the problematic process instances in which undesirable outcomes occur are known prior or easily detectable. For example, the problematic instances are those with outlier values or values bigger than a given threshold in one of the process features. However, on various occasions, using this approach, many related process improvement opportunities are not captured. To overcome this issue, we formulate the identification of process improvement areas as a context-sensitive anomaly detection problem. We define process improvement opportunities as a set of situations where the process performance is surprising. These surprising situations are situations in the process where the process performance is significantly different from the expected performance for other similar situations. Within the subgroups of situations, we can then further investigate the reason for the surprising behavior by applying root-cause analysis techniques using structural equation models. To evaluate the validity and relevance of the proposed approach, we have implemented and evaluated it on several real-life event logs. The experiments show that leveraging context awareness makes it possible to find additional process enhancement opportunities for different process perspectives in subsets of situations that were previously not captured.