[R] Bagged clustering and fuzzy c-means

[Friedrich.Leisch@ci.tuwien.ac.at]

>>>>> On Thu, 26 Jun 2003 14:43:41 +0100,
>>>>> Xu Yun (XY) wrote:

  > Dear All:
  > I'm a newbie to R and chemometrics.
  > Now I'm trying apply bclust on fuzzy c-means like this:
  >> bc1 <- bclust(iris[,1:4], 3, base.centers=20,iter.base=100,
  > base.method="cmeans")
  > Committee Member:
  > 1(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)Erro
  > r in bclust(iris[, 1:4], 3, base.centers = 20, iter.base = 100, base.method
  > = "cmeans") :
  >         Could not find valid cluster solution in 20 replications
  > I can't get any valid result with many parameter adjustments, such as
  > iter.base, base.centers etc. But I think fcm should return similar result
  > just like k-means (e.g. centers, cluster size) plus fuzzy membership
  > information. Can anyone explain this for me?

cmeans expects a matrix as input, iris[,1:4] is a data.frame.

bc1 <- bclust(as.matrix(iris[,1:4]), 3,
	base.centers=5,iter.base=100,base.method="cmeans")

works for me.

But I agree that this behaviour is not desirable, I'll add an
x=as.matrix(x) at the beginning of both bclust and cmeans


  > Besides, I'm not quite understand the meaning of "bootstrap". In my view, it
  > might means "independent", am I correct?

No, a bootstrap sample is a sample drawn from the empirical
distribution of a data set, i.e., drawing with replacement from the
original data.

There are heaps of books explaining what "bootstrapping" is, simply
search for books with "bootstrap" in the title in your library :-)

Best,

-- 
-------------------------------------------------------------------
                        Friedrich Leisch 
Institut für Statistik                     Tel: (+43 1) 58801 10715
Technische Universität Wien                Fax: (+43 1) 58801 10798
Wiedner Hauptstraße 8-10/1071      Friedrich.Leisch at ci.tuwien.ac.at
A-1040 Wien, Austria             http://www.ci.tuwien.ac.at/~leisch