

   CClluusstteerriinngg LLaarrggee AApppplliiccaattiioonnss

        clara(x, k, metric="euclidean", stand=F, samples=5, sampsize=40 + 2 * k)

   AArrgguummeennttss::

          x: data matrix or dataframe, each row corresponds to
             an observation, and each column corresponds to a
             variable. All variables must be numeric.  Missing
             values (NAs) are allowed.

          k: integer, the number of clusters.

     metric: character string specifying the metric to be used
             for calculating dissimilarities between objects.
             The currently available options are "euclidean"
             and "manhattan".  Euclidean distances are root
             sum-of-squares of differences, and manhattan dis-
             tances are the sum of absolute differences.

      stand: logical flag: if TRUE, then the measurements in
             `x' are standardized before calculating the dis-
             similarities. Measurements are standardized for
             each variable (column), by subtracting the vari-
             able's mean value and dividing by the variable's
             mean absolute deviation.

    samples: integer, number of samples to be drawn from the
             dataset.

   sampsize: integer, number of objects in each sample. `samp-
             size' should be higher than the number of clusters
             (`k') and at most the number of objects
             (nrow(`x')).

   DDeessccrriippttiioonn::

        `clara' is fully described in chapter 3 of Kaufman and
        Rousseeuw (1990).  Compared to other partitioning meth-
        ods such as `pam', it can deal with much larger
        datasets. Internally, this is achieved by considering
        sub-datasets of fixed size, so that the time and stor-
        age requirements become linear in nrow(`x') rather than
        quadratic.

        Each sub-dataset is partitioned into `k' clusters using
        the same algorithm as in the `pam' function.  Once `k'
        representative objects have been selected from the sub-
        dataset, each object of the entire dataset is assigned
        to the nearest medoid.  The sum of the dissimilarities
        of the objects to their closest medoid, is used as a
        measure of the quality of the clustering. The sub-
        dataset for which the sum is minimal, is retained.  A
        further analysis is carried out on the final partition.
        Each sub-dataset is forced to contain the medoids
        obtained from the best sub-dataset until then.  Ran-
        domly drawn objects are added to this set until `samp-
        size' has been reached.

   VVaalluuee::

        an object of class `"clara"' representing the cluster-
        ing.  See clara.object for details.

   BBAACCKKGGRROOUUNNDD::

        Cluster analysis divides a dataset into groups (clus-
        ters) of objects that are similar to each other. Parti-
        tioning methods like `pam', `clara', and `fanny'
        require that the number of clusters be given by the
        user.  Hierarchical methods like `agnes', `diana', and
        `mona' construct a hierarchy of clusterings, with the
        number of clusters ranging from one to the number of
        objects.

   NNOOTTEE::

        For small datasets (say with fewer than 200 observa-
        tions), the function `pam' can be used directly.

   RReeffeerreenncceess::

        Kaufman, L. and Rousseeuw, P.J. (1990). Finding Groups
        in Data: An Introduction to Cluster Analysis. Wiley,
        New York.

   SSeeee AAllssoo::

        `clara.object', `partition.object', `pam', `plot.parti-
        tion'.

   EExxaammpplleess::

                  #generate 500 objects, divided into 2 clusters.
        x <- y_rbind(cbind(rnorm(200,0,8),rnorm(200,0,8)),
                     cbind(rnorm(300,50,8),rnorm(300,50,8)))

        clarax <- clara(x, 2)
        clarax
        clarax$clusinfo
        plot(clarax)

