next up previous contents
Next: Set-relational operations for classes Up: Data reorganization Previous: Finding the indexes of

Merging a grouping and a classified data

 

mergecld Merge a classified data with groups
-c1 <groups> classified data for defining the groups of the classes in <cldata>
-c2 <cldata> classified data to be grouped
-cout <trgcldata> result classified data

This command merges classes of <cldata> according to the class groups defined in classified data <groups>. In this case, the identifiers in <groups> do not refer to actual data records. Instead, they refer to the classes in <cldata>. Resulting classes are named according to the classes in <groups>.

Example (ex3.1): In the case of SOMs, merging means that data records, which are divided between the neurons, are collected to larger classes. In other words, the indexes of the data records chosen to the neurons belonging to some neuron group are collected to the same class in the classified data <trgcldata>.

...
NDA> somtr -d boston -sout som1 -cout cld1 -l 4
...
NDA> mkgrp grp1 -s som1
...
# Grouping with a graphical tool -> groups are stored in the
# classified data "som1Grp". SOM classification for data records
# is stored in "cld1".
NDA> mergecld -c1 som1Grp -c2 cld1 -cout classif
# To utilize the classification "classif":
NDA> bin -d boston -c classif -dout data2
NDA> pickrec -d boston -c classif -dout data3

figure1027



Anssi Lensu
Wed Oct 6 12:57:48 EET DST 1999