Question

this is the sas code which i want to replicate in R,

proc fastclus data = in.stores_standard
maxclusters = 20
outseed= in.out_seed
maxiter = 1000
converge = 0 
strict=5.0; 
var storesize sales_per_sqft sales_per_visits tothhsinta;
id store_nbr;
run;

my attempt:

library(amap)
set.seed(1)
kmeans_object=Kmeans(stores_standard, 20, iter.max = 1000, nstart = 1, method = c("euclidean"))
p=do.call(rbind, kmeans_object)

What am unable to achieve:
1) run kmeans on these parameters only: storesize,sales_per_sqft,sales_per_visits, tothhsinta

2) id on store_nbr

3) outseed function in R

Thanks!

Answer 1

1) is quite easy:

want <- c("storesize", "sales_per_sqft", "sales_per_visits", "tothhsinta")
Kmeans(stores_standard[, want], 20, iter.max = 1000, nstart = 1,
       method = c("euclidean"))

For 2)

 ## a 2-dimensional example from ?Kmeans
 x <- rbind(matrix(rnorm(100, sd = 0.3), ncol = 2),
            matrix(rnorm(100, mean = 1, sd = 0.3), ncol = 2))
 colnames(x) <- c("x", "y")
 cl <- Kmeans(x, 2)

Now look at cl:

R> str(cl)
List of 4
 $ cluster : int [1:100] 2 2 2 2 2 2 2 2 2 2 ...
 $ centers : num [1:2, 1:2] 1.0245 -0.017 1.0346 0.0375
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : chr [1:2] "1" "2"
  .. ..$ : chr [1:2] "x" "y"
 $ withinss: num [1:2] 0.00847 0.22549
 $ size    : int [1:2] 50 50
 - attr(*, "class")= chr "kmeans"

The cluster component of the list contains the assigned cluster ID. These are in the same order as the samples in the input data. If you want to assign the cluster component as a column in the input data we’d then do:

R> x <- cbind(x, Cluster = cl$cluster)
R> head(x)
               x            y Cluster
[1,] -0.24251497  0.532012889       2
[2,]  0.10957740  0.225168920       2
[3,] -0.35563544 -0.428798979       2
[4,] -0.41251306  0.529953489       2
[5,] -0.61212001 -0.003443993       2
[6,]  0.04435213  0.086595025       2

For your data do:

stores_standard <- cbind(stores_standard, Cluster = kmeans_object$cluster)

As for 3, that doesn’t appear possible with kmeans() in standard R nor Kmeans() in package amap.