clcomb | Combine two classifications into one |
-d <indata> | original data frame |
-c1 <cldata1> | keys corresponding to the data 1 |
-c2 <cldata2> | keys corresponding to the data 2 |
{-cout <cldataout> | | combined classified data |
-dout <dataout>} | output data |
[-lout <lossout>] | output frame name for the number of unclassified records |
[-bin] | binary output instead of record counts |
This operation combines two classifications into one. The resulting classes are created according to <cldata2>, but data record indexes are replaced with the class numbers of <cldata1>. If some data record that has been identified in <cldata2> does not belong to any class of <cldata1>, it is recorded in <lossout>, which is a frame containing the number of unclassified data records for each class of <cldata2>. The output data frame contains the number of data records belonging to a certain class of <cldata2> (field) and <cldata1> (record number). Binary output can be requested with -bin. Binary output merely indicates whether some feature of <cldata2> exists for a class of <cldata1>.
Example: To collect all neuron IDs, which have captured a certain group of data records, following commands can be issued:
# Load data and train a TS-SOM NDA> load t.dat -n t NDA> select opetus -d t NDA> rm -fr opetus opetus.stid NDA> rm -fr opetus opetus.exid NDA> somtr -d opetus -sout som -cout cldata -l 6 ... NDA> somlayer -s som -fout soml # Select two fields representing unique ID of groups of data # records and create a classification, in which each class # represents one group NDA> select unifld -f t.stid t.exid NDA> uniq -d unifld -cout unicld # Select one layer of SOM classification NDA> selcld -c cldata -cout cldl5 -expr 'soml' = 5; NDA> clcomb -c1 unicld -c2 cldl5 -d t -dout hitcounts # The resulting frame contains the number of data records # having a certain combination of classifications # Fields are named as neuron_...