Advanced Methods for the Analysis of Semiconductor Manufacturing Process Data - Advanced Techniques in Knowledge Discovery and Data Mining - page 65

Fig. 2.22. Visualization of selected SPLIT6.
the method could be applied to create a starting solution for a higher-order
method in a hierarchical approach. Such a hierarchical approach is considered
very promising for future work.
The most meaningful result with regard to identified underlying physical
and technological evidence was achieved by the most recent FS variant, em-
ploying SBS and q ci for SPLIT6 as well as SPLIT3. Fifteen variables have
been selected (see Table 2.4), and a feature space with compact and well-
separated class regions is obtained by this selection. Figure 2.23 shows the
resulting projection of the 15-dimensional data of SPLIT3, which is definitely
superior to the result obtained for q si application. Regarding the underlying
physical meaning of the variables, the validity and significance of this selec-
tion is underpinned. Table 2.5 explains the meaning of the selected variables
for the regarded submicron CMOS process.