2.1 Structural improvement of model
There is no theory or formula to calibrate R. Many
models have been proposed to avoid the improper
selection of R. The PPDC model proposed by Chen
et al.[3] is the most widely used clustering model. The
model was the first to combine dynamic cluster thought
and projection pursuit. It avoided the issue of choosing
R effectively. In addition, it proposed a new approach
to assigning weight, based on the decision-makers’
preferences[8].
These models offer some help to users processing
high-dimensional clustering problems. Moreover, with
so many improvement models, it is hard for users to
determine which is most appropriate. As mentioned in
Ref. [9], existing models are still not good enough.