Page 325 -

12.2 Approach 307

Fig. 12.6 Filtering the event log before process discovery: (a) selecting activities that occur for at
least 5% of all 2765 patients, (b) selecting the top 80% of the 74 activities conducted by employees
of the housing agency, (c) C-net discovered based on a filtered log (the event log of the housing
agency after removing the activities occurring for less than 50% of the units)

cases in terms of a vector of variables (the features). By using a hierarchical clus-
tering technique as shown in Fig. 12.7, one can view the same process at multiple
levels. Cutting the dendrogram close to the root results in a few more complex mod-
els. Cutting the dendrogram closer to the leaves of the tree results in many simple
models.
In the next chapter, we describe an alternative way to simplify process models.
In contrast to filtering, simplification and abstraction techniques are directly applied
to the process graph. This so-called fuzzy mining approach views process models