Agfdhyk

After running k-means() on using different numbers of k = [2,3,4,5] on the iris dataset using the map() function, I would like to interpret the results for different k using a predefined function.

Below is my attempt:

library(dplyr)

library(purrr)



cluster_assignment <- map(2:5, function(k){

  result <- kmeans((x = iris[-5] %>%

                          scale()),

                    centers = k)



  # # return results to a list

  x <- list(result$cluster,

            result$tot.withinss,

            result$centers,

            result$size)

})





# assign cluster results back to the iris dataset

a <- map_dfc(cluster_assignment, 1)

colnames(a) <- paste0("result_", 2:5, "_cl")

iris <- bind_cols(iris, a)



> head(iris)

  Sepal.Length Sepal.Width Petal.Length Petal.Width Species result_2_cl result_3_cl result_4_cl result_5_cl

1          5.1         3.5          1.4         0.2  setosa           2           2           3           3

2          4.9         3.0          1.4         0.2  setosa           2           1           3           2

3          4.7         3.2          1.3         0.2  setosa           2           1           3           2

4          4.6         3.1          1.5         0.2  setosa           2           1           3           2

5          5.0         3.6          1.4         0.2  setosa           2           2           3           3

6          5.4         3.9          1.7         0.4  setosa           2           2           3           5

Now, I would apply a predefined function cluster_result2 to the newly assigned columns, i.e "result_2_cl", "result_3_cl", "result_4_cl", "result_5_cl"

# predefined function

cluster_result2 <- function(x, ...){

 x %>%

   group_by_(...) %>%

   summarise(size = n(),

             mean_spl = mean(Sepal.Length))

}



# tried this method, but did not get the expected output

map(iris[, colnames(a)], ~ cluster_result2(iris, .x))

How can I achieve this using the tidyverse approach? I found a very similar approach here, but couldn't get the expected output.

The expected output will be similar to the ones below, except they are stored in a nested list/dataframe:

> cluster_result2(iris, colnames(a)[1])

# A tibble: 2 x 3

  result_2_cl  size mean_spl

        <int> <int>    <dbl>

1           1   100     6.26

2           2    50     5.01

> cluster_result2(iris, colnames(a)[2])

# A tibble: 3 x 3

  result_3_cl  size mean_spl

        <int> <int>    <dbl>

1           1    21     4.75

2           2    33     5.17

3           3    96     6.31



  > cluster_result2(iris, colnames(a)[3])

# A tibble: 4 x 3

  result_4_cl  size mean_spl

        <int> <int>    <dbl>

1           1    29     7.00

2           2    50     6.14

3           3    49     5.02

4           4    22     5.50



> cluster_result2(iris, colnames(a)[4])

# A tibble: 5 x 3

  result_5_cl  size mean_spl

        <int> <int>    <dbl>

1           1    47     6.78

2           2    17     4.69

3           3    26     5.07

4           4    53     5.80

5           5     7     5.53

Appreciate your answers!

We can use group_by_at instead of group_by_ (it is deprecated). Here, we need to loop through the column names of 'a' instead of the columns of 'iris'

library(tidyverse)

map(colnames(a), ~ cluster_result2(iris, .x))

Or without using the ~, specify the 'x' parameter as 'iris'

map(colnames(a), cluster_result2, x = iris)

#[[1]]

# A tibble: 2 x 3

#  result_2_cl  size mean_spl

#        <int> <int>    <dbl>

#1           1    50     5.01

#2           2   100     6.26



#[[2]]

# A tibble: 3 x 3

#  result_3_cl  size mean_spl

#        <int> <int>    <dbl>

#1           1    47     6.78

#2           2    53     5.80

#3           3    50     5.01



#[[3]]

# A tibble: 4 x 3

#  result_4_cl  size mean_spl

#        <int> <int>    <dbl>

#1           1    50     6.14

#2           2    22     5.50

#3           3    29     7.00

#4           4    49     5.02



#[[4]]

# A tibble: 5 x 3

#  result_5_cl  size mean_spl

#        <int> <int>    <dbl>

#1           1    16     5.32

#2           2    29     7.00

#3           3    23     5.55

#4           4    34     4.86

#5           5    48     6.16

-checking with the output of function individually applied to columns

cluster_result2(iris, colnames(a)[4])

# A tibble: 5 x 3

#  result_5_cl  size mean_spl

#        <int> <int>    <dbl>

#1           1    16     5.32

#2           2    29     7.00

#3           3    23     5.55

#4           4    34     4.86

#5           5    48     6.16

NOTE: The output will be slightly different due to the randomness