| traj | Create or set a trajectory |
| <graphic> | name of the graphic structure |
| -n <name> | name for the trajectory |
| -min <min> | starting index of data |
| -max <max> | ending index of data |
| -co <color> | color for the trajectory |
| [-step <step>] | step between the labels |
| [-f <field>] | field for labelling. If omitted, indexes are used as labels |
| [-c <cldata>] | another classified data |
| [-d <ddata>] | distance data, which describes the distances between data records and their BMU neurons |
| traj <graphic> -n <name> -rm | Remove trajectory <name> |
This operation creates a trajectory onto a TS-SOM layer. Data records (indexes from <min> to <max>) are placed to neurons according to the SOM classification. If a classified data has been connected to a graphic structure with setgcld (see section 7.1.1), it defines the mapping between neurons and data records. However, one can override this by specifying a classified data with flag -c. The default way to locate a trajectory is to set trajectory points in the centroid of the subgraphs. You can adjust the location of trajectory points by specifying a distance data <ddata>. This distance data can be calculated with somproj (see section 5.1.9 and the second example).
Example (ex7.7): The following command sequence creates a graphic structure called grp1 and connects a classified data to it. Then, it creates a trajectory, which is labelled with data record indexes, which is shown for every 20th data record. Note that the trajectory depends on the layer of the TS-SOM, and it will be updated after setting the current layer.
... NDA> mkgrp grp1 -s som1 NDA> setgcld grp1 grp1 -c cld1 NDA> layer grp1 -l 3 NDA> traj grp1 -n tr1 -co blue -min 0 -max 300 -step 20
Example (ex7.8): The previous example fixed the trajectory lines to the centroids of the neurons. In this example, we compute a distance data, which describes the distances between data records and their BMU neurons. These distances are used to move the endpoints of trajectory lines from the centroids of the neurons.
Also the classified data alone can be used to map trajectories from distinct data sets to the same SOM. Thus, the projection data (here dmat) is not necessary for this command.
NDA> somproj -s som1 -w som1_W -d predata -c c1 -dout dmat
NDA> mkgrp grp1 -s som1
NDA> setgcld grp1 grp1 -c 1
NDA> layer grp1 -l 3
NDA> traj grp1 -n tr1 -co red -min 0 -max 300 -step 20 -c c1
-d dmat
