Advanced Methods for the Analysis of Semiconductor Manufacturing Process Data - Advanced Techniques in Knowledge Discovery and Data Mining - page 70

Fig. 2.27. Illustration of enhanced visualization features by the adapted AN.
nent and the related dimensionality-reduction methods, which also can be of
assistance to cluster, select, and rank features or measurement parameters.
2.6 Conclusions
The presented work contributes to the industrial application of advanced
soft-computing methods in the field of semiconductor manufacturing pro-
cess data analysis. In particular, fast and e cient methods for multivariate
data-dimensionality reduction, including automatic methods for parameter or
parameter group saliency detection, and interactive visualization have been
investigated in this first feasibility study.
Already the least complex and therefore most computationally inexpen-
sive visualization methods allow significant insight into the structure of the
data. Complemented by an interactive feature-selection tool, these visualiza-
tion methods represent a powerful addition to the standard statistical analysis
that is usually performed. Online visualization of the process trajectory in
the multivariate space is also feasible by available fast methods for adding
new data vectors in an existing mapping [2.23].
Furthermore, the investigation of automatic feature-selection methods has
yielded very promising results. For instance, from the resulting projection,
the asymmetry of the split can clearly be observed, which is a very signifi-
cant achievement. Additionally, even in those cases where variables selected