somproj | Compute the distances between data records and their BMU neurons |
-s <som> | name of the TS-SOM structure |
[-w <wmatrix>] | frame containing the weight matrix |
-d <data> | data to be classified |
-c <cldata> | SOM classification |
-dout <dataout> | name for the resulting data |
This operation computes the distances between data records and their BMU neurons. The operation is executed for each layer of the TS-SOM. The distances are computed in the direction of each SOM dimension, and every dimension of each layer will have its own field, which is named layer_dimension. The dimensions are indexed starting with zero. The distances are scaled into range [-1,1].
Example (ex5.5): This example shows how distances can be used with trajectories. When the distance data is given to traj (see section 7.2.3), trajectory lines are not drawn between the centroids of the neurons. Instead, the endpoints are moved according to the distances.
NDA> somtr -d pre1 -sout s1 -cout c1 -l 6 ... NDA> somproj -s s1 -w s1_W -d pre1 -c c1 -dout dist1 NDA> clstat -c c1 -d pre1 -dout sta1 -avg -min -max -var -hits NDA> mkgrp win1 -s s1 NDA> setgdat win1 -d sta1 NDA> setgcld win1 -c c1 NDA> layer win1 -l 4 # Using the mapping with a trajectory NDA> traj win1 -n a -co red -c c1 -d dist1 -min 0 -max 3000 -c c1 -d dist1