SOM map - R - Kohonen pack - clustering

Question

I try SOM maps and clustering in R. I use this tutorial : https://www.r-bloggers.com/self-organising-maps-for-customer-segmentation-using-r/ ... SOM maps work fine, but when I try clustering this error :

> mydata <- som_model$codes 
> wss <- (nrow(mydata)-1)*sum(apply(mydata,2,var)) 
Error in apply(mydata, 2, var) : dim(X) must have a positive length.

My code :

 require(kohonen)
    data = matrix( 
    c(6, 6, 80, 280, 404, 0, 158, 158197, 158197233,
      6, 13, 85, 280, 402, 0, 160, 160197, 160197233,
      6, 13, 81, 283, 400, 0, 160, 160197, 160197233),    
     nrow=3,             
     ncol=9,              
     byrow = TRUE)    
    data_train <- data[, c(1,2,4,5,7,8,9)]
    data_train_matrix <- as.matrix(scale(data_train))
    som_grid <- somgrid(xdim = 2, ydim=1, topo="hexagonal")

    som_model <- som(data_train_matrix, 
      grid=som_grid, 
    rlen=500, 
     alpha=c(0.05,0.01), 
     keep.data = TRUE )



      #training proces
        plot(som_model, type="changes")
        #nodes 
        plot(som_model, type="count", main="Node Counts")
        #heatmap
        plot(som_model, type = "property", property = getCodes(som_model)[,4], main="Heat map - status")

mydata <- som_model$codes 
wss <- (nrow(mydata)-1)*sum(apply(mydata,2,var)) 
for (i in 2:15) {
  wss[i] <- sum(kmeans(mydata, centers=i)$withinss)
}
plot(wss)

    ## use hierarchical clustering to cluster the codebook vectors
    som_cluster <- cutree(hclust(dist(som_model$codes)), 6)
    # plot these results:
    plot(som_model, type="mapping", bgcol = pretty_palette[som_cluster], main = "Clusters") 
    add.cluster.boundaries(som_model, som_cluster)

It is write same way like tutorial, so how is possible tutorial works and this not ? I am new in R so I dont understood this error. I understood that there is probably problem with matrix, but how problem ?

score 3 · Accepted Answer · answered Mar 05 '18 at 13:04

There are several problems with your code. First of all you start with a very small data sample, you perform SOM on a 2 x 1 grid which outputs just 2 rows in som_model$codes and then you perform kmeans with up to 15 clusters. I will provide working code with the Sonar data set from library mlbench. I must add I have never used kohonen library or SOM in real data analyses.

library(mlbench)
library(kohonen)
data(Sonar) #somewhat bigger data example

data_train <- Sonar[, 1:60] #use first 60 columns

data_train_matrix <- as.matrix(scale(data_train)) #scale data

som_grid <- somgrid(xdim = 5, ydim = 5, topo = "hexagonal") #initialize a bigger grid

som_model <- som(data_train_matrix, 
                 grid = som_grid, 
                 rlen = 500, 
                 alpha = c(0.05,0.01), 
                 keep.data = TRUE )

plot(som_model, type = "changes")

mydata <- som_model$codes[[1]] #extract the matrix containing codebook vectors

wss <- (nrow(mydata)-1)*sum(apply(mydata,2,var)) 

for (i in 2:24) { #i must be less than 5*5 the grid size defined at the begining
  wss[i] <- sum(kmeans(mydata, centers = i)$withinss)
}

plot(wss, type = "l")

lets use 8 clusters to cut the three:

som_cluster <- cutree(hclust(dist(mydata)), k = 8)

plot(som_model, type="mapping", bgcol = som_cluster, main = "Clusters") 
add.cluster.boundaries(som_model, som_cluster)

Thank you. In question i have small size example matrix, but in reality i use matrixs of many sizes 3x3,5x5, 10x10. It didnt work on none of this. But is true that i dont realize how important is number of clusters. — caroline, Mar 05 '18 at 19:20

score 0 · Answer 2 · answered Mar 05 '18 at 13:07

0

The error you're getting is - I believe - because som_model$codes is a list so you can't use apply on it, so replace your definition of mydata with: mydata <- som_model$codes[[1]], as in just add [[1]] to the end

then it should work

answered Mar 05 '18 at 13:07

DS_UNI

2,600
2
11
22

SOM map - R - Kohonen pack - clustering

2 Answers2