Clustering Event Traces by Behavioral Similarity

The automated analysis of event logs in smart homes could provide an IT-aided support for a highly autonomous and age-appropriate life standard. The analysis of human behavior in the context of smart homes is, however, a challenging task. Humans behave according to the best practices and a single behavioral model is typically not sufficient to represent them all. In fact, existing process mining algorithms reportedly generate spaghetti models from event logs of flexible processes, which are largely incomprehensible. Therefore, this paper presents a novel approach for clustering event traces by their behavioral similarity, rather than deriving a unique process model encompassing all traces. In order to do this, two algorithms are introduced and we report the results of a preliminary evaluation demonstrating the efficacy of the approach.