Description Usage Arguments Value Note Author(s) References See Also Examples
This function outputs a dissimilarity matrix of dissimilarities between the rows a data matrix computed by the COSA 2 algorithm. It is assumed that users are familiar with the COSA paper(s) or the vignette that comes with the rCOSA package, see references below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
X |
input data.frame, or matrix object in numeric mode. COSA calculates the dissimilarities for the rows in X. |
lX |
either an integer or a vector with as much elements as columns in |
targ |
target values for computing targeted dissimilarities. The
|
targ2 |
the second target value when computing dual targeted dissimilarities. The |
knear |
size of number of objects in the near-neighborhoods which is used to calculate attribute weights for each object. By default |
xmiss |
numeric value for missings in the data, by default it is set to |
lambda |
multiple attribute clustering incentive parameter. By default, the regularization parameter lambda is set to equal |
qntls |
quantiles used for calculating high and/or low targets (ignored if |
wtcomb |
by default is set to 'each', meaning that the maximum of the weights of object |
relax |
the number with which the homotopy parameter eta should be incremented at each outer iteration (for more info see noit)
|
conv |
the convergence treshold that can reduce the maximum number of inner iterations.
|
niter |
the maximum number of inner iterations to stabilize the weights and dissimilarties given the homotopy parameter
|
noit |
the number of outer iterations ( make sure relax > 0 ) to transfer from the inverse exponential distance more closely to the sum of the weighted dissimilarities, obtained when a large enough number of outer iterations is chosen. Starting with the initial value of the homotopy parameter (equal to lambda) and using increments determined by relax one can calculate at what value the homotopy parameter will end.
|
stand |
equals |
pwr |
|
This function outputs a list that has as the first element the call, as the second element the dissimilarity matrix out$D
of class dist, and by default also the weights of class matrix, and, if crit
is set to TRUE
, the values of the tuning parameters are also given in the output.
The output dissimilarity matrix can be used as input to most dissimilarity based clustering procedures in R in the same manner as the output of the procedure dist
.
Maarten M.D. Kampert, Jacqueline J. Meulman, and Jerome H. Friedman.
Correspondence: mkampert@math.leidenuniv.nl
Friedman, J. H. and Meulman, J. J. (2004). Clustering objects on subsets of attributes.
URL: http://www-stat.stanford.edu/~jhf/ftp/cosa.pdf
Kampert, M.M., Meulman J.J., Friedman J.H. (2017). rCOSA: A Software Package for Clustering Objects on Subsets of Attributes
URL: https://link.springer.com/article/10.1007/s00357-017-9240-z
hierclust
, getclust
, and smacof
.
1 2 3 4 5 6 |
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.