There are two ways of doing this depending upon what you want. In the explorer on the preprocess tab, select the appropriate filter with the settings and then apply it. You can save the dataset using the ``Save'' button.

  1. weka.filters.unsupervised.attribute.AddCluster - A filter that adds a new nominal attribute representing the cluster assigned to each instance by the specified clustering algorithm.
  2. weka.filters.unsupervised.attribute.ClusterMembership - A filter that uses a clusterer to obtain cluster membership values for each input instance and outputs them as new instances. The clusterer needs to be a density-based clusterer. If a (nominal) class is set, then the clusterer will be run individually for each class.

If you want to do so from the command line then one way to do so is use the above filters:

java -cp weka.jar; weka.filters.unsupervised.attribute.AddCluster -W 
weka.clusterers.SimpleKMeans -i data/iris.arff

The other alternative is to use weka.clusterers.ClusterEvaluation and its 'p' option:

java -cp weka.jar; weka.clusterers.ClusterEvaluation 
weka.clusterers.SimpleKMeans -t data/iris.arff -p 1-5

With the 'p' option you can specify the range of instance values which should be output. Use '0' for none.

Source: Weka FAQ