||Math -> Cluster Analysis -> kMeans Clustering...
The command Math/Cluster Analysis/kMeans Clustering... (toolbar button ) allows to perform a k-means cluster analysis. The resulting cluster numbers can be assigned to the class numbers of the dataset.
||Please follow these steps to perform a k-means cluster analysis:
- Select the variables by clicking the list of descriptors. This opens the variable selection dialog.
- Set the parameters of the analysis, i.e. the number of expected clusters and the scaling of the data.
- Click the button "Calculate"
- Inspect the results and click "Assign Classes" to copy the found clusters to the class numbers of the dataset.
- In the case you want to revert the class assignment click the undo button
||The results of the k-means procedure depend on a random initialisation, thus the may differ from run to run. If you want to find an optimum solution click the button "Optimize". This will repeat the algorithm N times and will return the best result of all runs. The optimum is condidered to be the solution with the minimum intra-cluster distance.