Expert-driven trace clustering with instance-level constraints

Within the field of process mining, several different trace clustering approaches exist for partitioning traces or instances into similar groups. Typically, this is based on certain patterns similarity between traces, driven by discovery a model each cluster. The main drawback these techniques, however, that their solutions are usually hard to evaluate justify domain experts. In paper, we present two constrained techniques capable leverage expert knowledge in form instance-level constraints. an extensive experimental evaluation using real-life datasets, show our novel indeed producing more justifiable without substantial negative impact quality.