1 Reply Latest reply: Jun 20, 2012 11:17 AM by 400983 RSS

    Clustering in SQL Developer

    944112
      Hello All,

      I have a couple of questionings regarding cluster build.

      1) Is it possible to have a min set amount per leaf/cluster? (As in I don't want to create a cluster with just 1 person it doesn't really tell me anything)

      2) Is it possible to select K-means by the median rather than the centroid?

      3) Is it possible to view the probability formula for the k-means and O-Cluster?

      Thanks,

      Alan
        • 1. Re: Clustering in SQL Developer
          400983
          Hi Alan,

          Please see responses inline.

          Thanks,
          Boriana

          1) Is it possible to have a min set amount per leaf/cluster? (As in I don't want to create a cluster with just 1 person it doesn't really tell me anything)
          Setting a min count for a leaf is not supported. However, to avoid very unbalanced clusters, you can change the split criterion setting to size. This would produce balanced clusters.

          2) Is it possible to select K-means by the median rather than the centroid?
          K-medoid is currently not supported.

          3) Is it possible to view the probability formula for the k-means and O-Cluster?
          For k-Means, the probability is based on using Gaussians with the centroids as means and standard deviations based on the dispersion.
          For O-Cluster, the probability is based on an internally computed Baysian model.