(noise) Let CL={C1 ,. . .,Ck} be a clustering of the database D with respect to
NPred, MinWeight. Then we define the noise in D as the set of objects in the database D not belonging to any density-connected set Ci , i.e. noiseCL = D (C1 ∪ . . . ∪ Ck).
There are other possibilities to define a clustering based on definition 6. But this simple notion
of a clustering has the nice property that two clusters can at most overlap in objects which are
border objects in both clusters. Figure 4 illustrates the overlap of two clusters using cardinality
and MinCard = 4.